# Gloo Forge Documentation

> Forge is a hybrid workforce orchestration platform: humans and AI agents collaborating in production through governed workflows, durable execution, real-time chat, and shared dev environments.

This file is the machine-readable index for the Forge docs site, following the [llms.txt convention](https://llmstxt.org/). The companion file `llms-full.txt` contains the full corpus inline for one-shot ingestion.

_Generated by `scripts/generate-llms-txt.mjs`. 253 pages._

## Forge Console

- [Forge Console](https://docs.gloo.com/forge/console): Forge Console is the operator dashboard for Gloo Forge. Home gives operators an at-a-glance view of the whole platform: a customizable widget grid driven by reactive Convex subscriptions.
- [Analytics — The Ledger](https://docs.gloo.com/forge/console/analytics): The Ledger — cost, budgets, throughput, reliability, agent performance, and alerts. Six aggregate surfaces over one rollup primitive.
- [Agent Performance](https://docs.gloo.com/forge/console/analytics/agent-performance): Per-agent performance roll-up — dispatches, success rate, average cost, and the four-tab Synthesis / Builders / Architect sub-views.
- [Alerts](https://docs.gloo.com/forge/console/analytics/alerts): Workspace alert configuration — the alert rules that turn observable thresholds into operator notifications.
- [Architect Health](https://docs.gloo.com/forge/console/analytics/architect-health): Architect-specific reliability surface — workflow-graph authoring, debug consultations, and impact-analysis sessions.
- [Budgets](https://docs.gloo.com/forge/console/analytics/budgets): Budget envelope management — daily and monthly caps per factory, team, workflow, or agent.
- [Cost](https://docs.gloo.com/forge/console/analytics/cost): Token, compute, and platform-API spend rolled up across workflows, agents, factories, and teams.
- [Reliability](https://docs.gloo.com/forge/console/analytics/reliability): Run success rates, error budgets, and incident counts — sliced by workflow, agent, and time window.
- [Throughput](https://docs.gloo.com/forge/console/analytics/throughput): Workflow and agent throughput — runs per hour, tokens per hour, sandbox-minutes per hour — over time and by dimension.
- [Chat — The Bell](https://docs.gloo.com/forge/console/chat): The operator experience for channels, sessions, and real-time agent conversations in Forge Console. Two surfaces: Personal Agent and Conversations.
- [Conversations](https://docs.gloo.com/forge/console/chat/conversations): The full channel directory — every factory channel, project channel, team channel, direct message, and personal channel you have access to.
- [Personal Agent](https://docs.gloo.com/forge/console/chat/personal-agent): Your private 1:1 chat with the Personal Agent — the operator-bound agent that owns your roadmap, briefings, and personal task queue.
- [Observe — The Watchtower](https://docs.gloo.com/forge/console/observe): Read-only telemetry for Forge: events, run history, logs, traces, and LLM routing. Five passive surfaces. Active alerting lives in Analytics.
- [Events](https://docs.gloo.com/forge/console/observe/events): The unified platform event stream — every governance decision, lifecycle transition, dispatch, and error in one filterable feed.
- [LLM Routing](https://docs.gloo.com/forge/console/observe/llm-routing): Per-dispatch view of which model handled each agent turn, why it was chosen, and how the model-router fallback chain resolved.
- [Logs](https://docs.gloo.com/forge/console/observe/logs): Structured log search across every workflow run, agent dispatch, and Trust Fabric gate decision.
- [Run History](https://docs.gloo.com/forge/console/observe/run-history): Historical run feed — every workflow run with its outcome, duration, cost, and the worker who initiated it.
- [Traces](https://docs.gloo.com/forge/console/observe/traces): Distributed traces — every Convex action, agent turn, and tool call captured as a span in OpenTelemetry-style waterfall view.
- [Planning — The Blueprint](https://docs.gloo.com/forge/console/planning): Strategic planning surface — Roadmaps, Goals, Blockers, and Critical Path. Where quarterly delivery is mapped, sequenced, and unblocked.
- [Blockers](https://docs.gloo.com/forge/console/planning/blockers): Cross-roadmap blocker board — every flagged blocker, the milestones it impacts, and the assigned resolver.
- [Critical Path](https://docs.gloo.com/forge/console/planning/critical-path): The longest dependency chain across active roadmaps — the milestones that, if delayed, push the planned completion date.
- [Goals](https://docs.gloo.com/forge/console/planning/goals): Goal tree — strategic objectives broken down into measurable outcomes, with progress tracked against milestone delivery.
- [Roadmaps](https://docs.gloo.com/forge/console/planning/roadmaps): Quarterly roadmap view — time-bounded planning containers with workstream swim-lanes and per-milestone Gantt bars.
- [Platform — The Foundry](https://docs.gloo.com/forge/console/platform): Workspace administration. Models, skills, tools, commands, procedures, synthesis, judges, schedules, capabilities, and settings.
- [Capabilities](https://docs.gloo.com/forge/console/platform/capabilities): The capability provider registry — every platform capability domain, the foundation provider, and the chain of operator-selectable external providers.
- [Commands](https://docs.gloo.com/forge/console/platform/commands): Browse the typed command catalog and audit free-form exec invocations from the Foundry.
- [Judges](https://docs.gloo.com/forge/console/platform/judges): Inspect Judge agents, review calibration trends, override verdicts, manage the provisional/approved/deprecated tier lifecycle, and triage admissions decisions.
- [Models](https://docs.gloo.com/forge/console/platform/models): The model catalog and the per-provider routing chain — what models the platform can dispatch to, and how the model router resolves a request.
- [Procedures](https://docs.gloo.com/forge/console/platform/procedures): Browse, inspect, and audit sandbox-executed Procedures — typed, deterministic units of model-free work that act as workflow graph nodes and agent tools.
- [Schedules](https://docs.gloo.com/forge/console/platform/schedules): See every scheduled job on the platform — recurring crons (the load-bearing layer), operator-defined workflow triggers, and pending one-off jobs.
- [Settings](https://docs.gloo.com/forge/console/platform/settings): Workspace-level settings — feature flags, member admin, identity providers, chat configuration, webhooks, and external apps.
- [Skills](https://docs.gloo.com/forge/console/platform/skills): The skill catalog — packaged instruction sets that compose into agent prompts at dispatch time.
- [Synthesis](https://docs.gloo.com/forge/console/platform/synthesis): The synthesis pipeline — agent self-improvement runs that propose and test changes to prompts, skills, and procedures.
- [Tools](https://docs.gloo.com/forge/console/platform/tools): The tool registry — every tool the platform can grant to an agent, organized into tool sets, with governance and audit per-tool.
- [Production — The Flame](https://docs.gloo.com/forge/console/production): Where work happens. Factories, workflows, runs, approvals, delegated work, and sandboxes — the six operational surfaces in the Forge Console.
- [Approvals](https://docs.gloo.com/forge/console/production/approvals): The HITL approval queue — every pending decision raised by Trust Fabric `hold` outcomes, awaiting an authorized operator.
- [Delegated Work](https://docs.gloo.com/forge/console/production/delegated-work): The operator queue for work an agent has handed back to a human — A2A consultations, subagent waits, background tasks awaiting input.
- [Factories](https://docs.gloo.com/forge/console/production/factories): Stand up and manage Factories — autonomous production units pairing a Factory Manager agent with a team, budget, channels, and dev environments.
- [Runs](https://docs.gloo.com/forge/console/production/runs): Cross-workflow run feed — every active and recent run, with live DAG, timeline, logs, artifacts, and cost.
- [Sandboxes](https://docs.gloo.com/forge/console/production/sandboxes): Every Vercel Sandbox VM spawned by `runInSandbox` — live state, replay, diff, and lineage views.
- [Workflows](https://docs.gloo.com/forge/console/production/workflows): Workflow catalog — browse, search, pin, run, and edit every workflow definition in the workspace.
- [Trust Fabric — The Covenant](https://docs.gloo.com/forge/console/trust-fabric): The operator surface for Forge's governance system. Sub-grouped by the four SATO pillars: Identity, Policies, Traceability, Controls.
- [Agent Trust](https://docs.gloo.com/forge/console/trust-fabric/agent-trust): Per-agent trust posture — autonomy tier, calibration band, identity attestation, and the active credentials each agent holds.
- [Compliance Metrics](https://docs.gloo.com/forge/console/trust-fabric/compliance-metrics): Workspace-level compliance roll-up — audit completeness, attestation coverage, policy review currency, and incident counts.
- [Credential Approvals](https://docs.gloo.com/forge/console/trust-fabric/credential-approvals): Queue for credential-issuance requests — every time an agent or worker asks for a new credential, the request lands here for review.
- [Environments](https://docs.gloo.com/forge/console/trust-fabric/environments): Per-environment trust posture and vault inventory — the vaults, secrets, and credentials each environment hosts.
- [Gates](https://docs.gloo.com/forge/console/trust-fabric/gates): The numbered Trust Fabric gate pipeline — per-gate health, recent decisions, and gate-level configuration.
- [Key Issuance](https://docs.gloo.com/forge/console/trust-fabric/key-issuance): Cryptographic-key lifecycle — issue, rotate, and revoke the signing and encryption keys workers present to downstream systems.
- [Kill Switches](https://docs.gloo.com/forge/console/trust-fabric/kill-switches): Operations Kill Switches — platform-wide and budget-scoped emergency controls. Admin-gated, written-reason required, audit-logged under category ops_kill_switch.
- [Policies](https://docs.gloo.com/forge/console/trust-fabric/policies): Workspace policy catalog — every gate-6 declarative policy, with its scope, predicate, and current activation state.
- [Protected Data](https://docs.gloo.com/forge/console/trust-fabric/protected-data): The protected-data classifier and policy surface — what counts as sensitive, who can read it, and how dispatches that touch it are gated.
- [Risk Matrix](https://docs.gloo.com/forge/console/trust-fabric/risk-matrix): The OG-05 entry point for the SATO governance program — every governed action calibrated on consequence × reversibility, with the resulting tier driving autonomy posture, approval thresholds, and Two-Agent override requirements.
- [Traceability Trail](https://docs.gloo.com/forge/console/trust-fabric/traceability): Append-only governance evidence — Decision Records, Execution Traces, and Overrides. Reconstruct what happened, in what order, at what cost, under whose authority.
- [Vendors](https://docs.gloo.com/forge/console/trust-fabric/vendors): External-vendor risk register — the trust posture for every external system the platform integrates with.
- [Workforce — The Guild](https://docs.gloo.com/forge/console/workforce): The unified org chart for humans and agents. Three surfaces — Organization, People, Agents — across one shared worker abstraction.
- [Agents](https://docs.gloo.com/forge/console/workforce/agents): Agent fleet — every `AGENT` worker, the agent definition each binds to, governance posture, and the dispatches each has handled.
- [Organization](https://docs.gloo.com/forge/console/workforce/org-chart): The org chart — the company / department / team / position hierarchy, with both human and agent workers as a single abstraction.
- [People](https://docs.gloo.com/forge/console/workforce/people): Human members directory — search, filter, and inspect every `EMPLOYEE` and `CONTRACTOR` in the workspace.

## Forge Platform

- [Forge](https://docs.gloo.com/forge): Forge is the production runtime for AI agents. Durable workflows, human approvals, governance, observability, and an operator console — in one platform.
- [Agent Composition](https://docs.gloo.com/forge/architecture/agent-composition): The composeAgentConfig pipeline — how Forge resolves agent definitions into ready-to-run configurations.
- [Convex Platform](https://docs.gloo.com/forge/architecture/convex-engine): How Convex serves as the orchestration brain — durable state, workflow graph orchestrator, Trust Fabric, and Convex-native agents.
- [Convex Infrastructure](https://docs.gloo.com/forge/architecture/convex-infrastructure): Concrete limits, decision rules, and operational constraints for working with Convex as the substrate beneath Forge. Read this when designing schemas, indexes, queries, or actions.
- [Convex-Native Agents](https://docs.gloo.com/forge/architecture/convex-native-agents): Building durable agents inside Convex: the composeAgentConfig pipeline, the agent inventory, tool authoring, and the @convex-dev/agent runtime.
- [Data Flow End-to-End](https://docs.gloo.com/forge/architecture/data-flow-end-to-end): Tracing an agent invocation from trigger through composition, governance, execution, and persistence — across chat, factory, and workflow surfaces.
- [Data Model](https://docs.gloo.com/forge/architecture/data-model): The 140+ table Convex schema organized by domain — workflows, agents, governance, identity, budgets, and chat.
- [Execution Model](https://docs.gloo.com/forge/architecture/execution-model): How Forge agents execute — Convex-native dispatch with optional Vercel Sandbox for isolated code.
- [Files and Bindings](https://docs.gloo.com/forge/architecture/files-and-bindings): The unified files/bindings model — putFile persists bytes and creates a binding, getFile reads them. Replaces the legacy writeArtifact / readArtifact / copyFileToSandbox / importFileFromSandbox quartet.
- [Approval Lifecycle](https://docs.gloo.com/forge/architecture/flows/approval-lifecycle): How human-in-the-loop approvals are created, escalated, resolved, and how the workflow resumes. Now unified under the humanInteractionRequests table — every prompt (approval, question, confirmation, form) shares the same fan-out, recall, and callback pipeline.
- [Governance Pipeline](https://docs.gloo.com/forge/architecture/flows/governance-pipeline): The fail-fast safety pipelines that evaluate workflow dispatch and launch requests before execution is permitted.
- [Observability Architecture](https://docs.gloo.com/forge/architecture/flows/observability-architecture): How Forge watches itself — the path from event emission through ingestion, trace context, Tier 2 rollups, and reactive surfaces.
- [Workflow Execution Flow](https://docs.gloo.com/forge/architecture/flows/workflow-execution): The end-to-end execution flow from run start through graph walk, dispatch, governance, and completion on the workflowV2 orchestrator.
- [Forge API](https://docs.gloo.com/forge/architecture/forge-api): The Convex API surface that powers both the Forge Console and external HTTP integrations — `convex/forgeApi/`, organized by resource domain.
- [Judges Engine](https://docs.gloo.com/forge/architecture/judges-engine): How Forge dispatches Judges and consumes their verdicts. The subject builder, the verdict-schema gate, panel consensus, metrics cron, and per-workspace settings.
- [LLM Routing Tier Abstraction](https://docs.gloo.com/forge/architecture/llm-routing): Categorical tiers replace hard-coded model strings. Authors declare a tier; the platform resolves (tier × provider × runtime conditions) into a concrete model at dispatch time.
- [Procedure Governance Gates](https://docs.gloo.com/forge/architecture/procedure-governance-gates): The Trust Fabric gate pipeline a procedure dispatch passes through. Capability allowlist, egress allowlist, schema preflight — the procedure-specific siblings to the convex-native agent pipeline.
- [Procedures Subsystem](https://docs.gloo.com/forge/architecture/procedures-subsystem): Registration → snapshot build → executor → capability proxy. How a procedure goes from a directory in the repo to a pinned image executing in Vercel Sandbox.
- [Sandbox-as-Substrate](https://docs.gloo.com/forge/architecture/sandbox-substrate): The Vercel Sandbox VM is the platform's code-execution substrate, not a node attribute. How it boots, how it warms, how it snapshots, and how dispatch threads identity, secrets, and policy into the VM.
- [Schema Patterns](https://docs.gloo.com/forge/architecture/schema-patterns): Convex schema design at scale: 87+ tables, index strategy, validator patterns, and migration awareness.
- [Slack Gateway](https://docs.gloo.com/forge/architecture/slack-gateway): The Convex-native relay that lets a single Slack app drive workflows running on any Forge deployment — self-describing action tokens, HMAC-signed envelopes, presentation receipts, and observability hooks.
- [JIT Synthesis Pipeline](https://docs.gloo.com/forge/architecture/synthesis-pipeline): How the Workflow Architect, the Builders, the Coherence Gate, and the Curator compose into a closed loop. Search-first reuse, dedup, workspace budget, sprawl prevention.
- [System Architecture](https://docs.gloo.com/forge/architecture/system-overview): The three-layer architecture of Gloo Forge — how Forge Console, Convex, and the agent execution environments fit together, plus the monorepo layout, tech stack, and design decisions engineers should know before contributing.
- [Trust Fabric Internals](https://docs.gloo.com/forge/architecture/trust-fabric-internals): Deep dive into the governance engine: gate composition, hooks, role resolution, capability lifecycle, presentation receipts, and how to extend the system.
- [Two Interfaces](https://docs.gloo.com/forge/architecture/two-interfaces): Forge exposes one platform through two interfaces. The reactive client serves apps inside the monorepo; the HTTP API serves everything outside. Both read from the same source of truth at `convex/forgeApi/`.
- [Vercel Sandbox](https://docs.gloo.com/forge/architecture/vercel-sandbox): Isolated VM execution via the runInSandbox tool — lifecycle, API, snapshotting, and when to use it.
- [Workflow Graph Model](https://docs.gloo.com/forge/architecture/workflow-graph-model): The workflowV2 engine — graph data model, node types, the orchestrator state machine, and concurrency control.
- [Build with Forge](https://docs.gloo.com/forge/build): How-to recipes for shaping the platform — defining agents, composing workflows, authoring skills, registering tools, wiring capability providers.
- [Author a Judge](https://docs.gloo.com/forge/build/author-a-judge): Define a rubric skill, register the Judge agent, calibrate against operator overrides. Authoring a Judge is authoring an agent — there is no Judge framework, only the JudgeVerdictV1 contract.
- [Author a Procedure](https://docs.gloo.com/forge/build/author-a-procedure): Write a Python Procedure that runs in a Vercel Sandbox — the forge.yaml manifest, SDK primitives, canonical wire shapes, file bindings, and the built-in procedure catalog.
- [Author a Skill](https://docs.gloo.com/forge/build/author-a-skill): Three paths into skillDefinitions — builtin (repo), workspace (Console), imported (marketplace) — plus when to clone, when to flag a skill on_demand, and what allowedTools really does.
- [Compose a Workflow](https://docs.gloo.com/forge/build/compose-a-workflow): Author a workflow as a directed graph — pick the right node types, wire typed I/O, choose validation strategies, set capability hints. The mental model and the decisions before publish.
- [Configure a Sandbox](https://docs.gloo.com/forge/build/configure-sandbox): When to add runInSandbox to an agent's tool set, how to declare per-tool command whitelists, and the common pitfalls (cold-start latency, snapshot reuse, credential injection).
- [Define an Agent](https://docs.gloo.com/forge/build/define-an-agent): How to author a Forge agent — the three definition layers, the dispatch pipeline you're feeding, and the decisions you make once vs the ones you make per node.
- [Register a Tool](https://docs.gloo.com/forge/build/register-a-tool): Add a new tool to the Forge tool registry — when to register vs invoke an existing tool, the createTool/createGovernedTool boundary, and how the tool reaches an agent at dispatch.
- [Set Up an External App](https://docs.gloo.com/forge/build/set-up-external-app): Connect Forge to an external service — Linear, GitHub, Slack, 1Password, Pipedream — through the Virtual Keys credential model and the capability provider chain.
- [Talk to Agents](https://docs.gloo.com/forge/build/talk-to-agents): Wire chat into your product — channels, sessions, dispatch modalities, relevancy gating, personal agents. The build-side surface for hybrid workforce conversations.
- [Use the Workflow Architect](https://docs.gloo.com/forge/build/use-workflow-architect): How to author workflows by talking to the Workflow Architect — a Forge-native agent that turns intent into a validated graph you review and ship.
- [Wire a Capability Provider](https://docs.gloo.com/forge/build/wire-capability-provider): Add a new external provider to an existing capability domain — or stand up a brand-new domain. The four-step pattern, the file layout, the env-config wiring, and the consumption path.
- [Changelog](https://docs.gloo.com/forge/changelog): Weekly cadence — what shipped, what moved, what to know. Linear-style entries with hero media, highlights, and plan tags.
- [Week of 2026-04-28 — Sandbox Substrate, RFC #675 RLS, Capabilities widening](https://docs.gloo.com/forge/changelog/entries/2026-04-28): The substrate week. Sandbox phases 1A–4, RLS on forgeApi, capabilities chains, HITL approvals.
- [Week of 2026-05-02 — Forge Docs Strategy lands end-to-end](https://docs.gloo.com/forge/changelog/entries/2026-05-02): Thirty-one PRs, seven phases, one week. Forge docs caught up to Forge.
- [Week of 2026-05-05 — Files-and-Bindings unified model](https://docs.gloo.com/forge/changelog/entries/2026-05-05): Files-and-Bindings narrow lands its public-API contract change: /api/v2/files now returns Id<files>.
- [Week of 2026-05-27 — Docs refresh: foundations, reference, and new concepts](https://docs.gloo.com/forge/changelog/entries/2026-05-27): Two-cut docs refresh — foundations & reference (PR 1) plus new concepts (PR 2: Procedures, Judges, JIT Synthesis, Workflow Architect, Personal Agents) and the use-case audit pass.
- [Architecture Migration History](https://docs.gloo.com/forge/changelog/migration-history): A migration guide for anyone who knew the old architecture — what was retired, what replaced it, and where to find things now.
- [Key Concepts](https://docs.gloo.com/forge/concepts): The mental model for Forge — four families of concepts that compose into the platform: workforce, production, trust fabric, and substrate.
- [Composable Capabilities](https://docs.gloo.com/forge/concepts/capabilities): How Forge stays vendor-neutral. A capability is a typed interface for an external dependency; every domain ships an internal Convex-native fallback plus optional external providers, swappable per deployment.
- [Glossary](https://docs.gloo.com/forge/concepts/glossary): Canonical definitions for every Forge concept, surface, and term — the single reference for what we mean.
- [JIT Capability Synthesis](https://docs.gloo.com/forge/concepts/jit-synthesis): Just-in-time creation of skills, agents, and procedures when the Workflow Architect needs a primitive that doesn't yet exist. Provisional-First Reuse, Coherence Gate, Curator sweep, workspace synthesis budget.
- [Judges](https://docs.gloo.com/forge/concepts/judges): Verdict-emitting agents that grade other agents' output, workflow runs, primitives, and synthesis proposals. First-class workers — same dispatch infra, scoped tools, standardized verdict schema.
- [Personal Agents](https://docs.gloo.com/forge/concepts/personal-agents): One agent per human, provisioned automatically. Acts under the human's principal identity. The default first contact for chat in Forge — the agent that knows you, your factories, and your in-flight work.
- [Procedures](https://docs.gloo.com/forge/concepts/procedures): Typed, sandbox-executed units of model-free work. Peers to agents in the workflow graph — same execution contract, no LLM in the loop.
- [Production Concepts](https://docs.gloo.com/forge/concepts/production): The structures that orchestrate work over time. Workflows, runs, tasks, sessions, projects — the graph entity model and the orchestrator that walks it.
- [Substrate Concepts](https://docs.gloo.com/forge/concepts/substrate): The engine. Convex as the source of truth and orchestrator, Vercel Sandbox for isolated code execution, capability providers for vendor neutrality, durable execution semantics, and the agent composition layer that ties them together.
- [Tool Registry](https://docs.gloo.com/forge/concepts/tool-registry): The single source of truth for tools in Forge — what an agent can call, how it's wrapped, and how every invocation is audited.
- [Trust Fabric Concepts](https://docs.gloo.com/forge/concepts/trust-fabric): The governance system. Safety gates, governance decisions, policies, approvals, budgets, autonomy rungs, kill switches, audit. The umbrella that constrains what workers are allowed to do.
- [Workflow Architect](https://docs.gloo.com/forge/concepts/workflow-architect): The conversational agent that turns operator intent into a workflow graph. Propose-plan / confirm handshake, narrowed tool surface during confirmation, Architect-Judge graded output.
- [Workforce Concepts](https://docs.gloo.com/forge/concepts/workforce): The primitives that answer who does the work. Workers, agents, factories, teams, identity claims, and the worker-abstraction model that lets humans and agents share governance, budgeting, and attribution.
- [Component Gallery](https://docs.gloo.com/forge/contributing/component-gallery): Live examples of every Forge documentation component, with the JSX you copy into your page.
- [Forge Docs Diagrams](https://docs.gloo.com/forge/contributing/diagrams): How Forge docs use diagrams. Which component for which intent. The bar new diagrams are graded against.
- [Engineering Principles](https://docs.gloo.com/forge/contributing/engineering-principles): The *why* behind every architectural decision in Forge. A portable philosophical reference for integrators building on the platform.
- [Engineering Standards](https://docs.gloo.com/forge/contributing/engineering-standards): The *how* — concrete house rules for working in and on Forge. Operational and specific.
- [Forge Docs Frontmatter Contract](https://docs.gloo.com/forge/contributing/frontmatter): The metadata every Forge doc page carries. What each field means, how to set it on a new page, when to update it, and which CI workflow enforces it.
- [Forge Docs Voice](https://docs.gloo.com/forge/contributing/voice): How Forge docs sound. The seven patterns the canonical pages use, the rules new pages are held to, and the bar to grade against.
- [Guided Tour](https://docs.gloo.com/forge/guided-tour): A conceptual walk through Forge — architecture, agents, skills, workflows, governance, and the API.
- [How Forge Thinks](https://docs.gloo.com/forge/how-forge-thinks): The five invariants that shape every Forge feature. Read this before reading anything else.
- [Builder Telemetry](https://docs.gloo.com/forge/internal/builder-telemetry): Per-turn telemetry + daily metrics rollups + drift alerts for Skill Builder and Agent Builder — internal infrastructure that powers the /analytics/builders dashboard.
- [Knowledge Base](https://docs.gloo.com/forge/knowledge-base): Recipes, patterns, and worked examples that span multiple Forge capabilities — the Monday-morning answers to 'how do I X' that don't fit a how-to guide or a reference page.
- [Add a Workflow with HITL Approval](https://docs.gloo.com/forge/knowledge-base/add-workflow-with-hitl-approval): End-to-end recipe — compose a workflow graph with a human-in-the-loop approval node, attach a budget envelope, and watch a real run pause, resume, and complete with full audit.
- [Replace a Model on a Live Agent](https://docs.gloo.com/forge/knowledge-base/replace-model-on-live-agent): End-to-end recipe — register a new model in the catalog, swap an active agent's preferred model, and roll the change to production with no in-flight runs lost and full audit of the swap itself.
- [Ship a Scheduled, Recurring Workflow](https://docs.gloo.com/forge/knowledge-base/scheduled-recurring-workflow): End-to-end recipe — turn a working workflow definition into a cron-driven recurring run with envelope budgets, attention items on failure, and a kill-switch the operator can pull.
- [Ship an Audited Agent](https://docs.gloo.com/forge/knowledge-base/ship-an-audited-agent): End-to-end recipe — from a fresh agent definition to a Convex-native dispatch with every tool call, every governance decision, and every token spend recorded in the audit trail.
- [Use Cases](https://docs.gloo.com/forge/knowledge-base/use-cases): Real-world scenarios that demonstrate Forge capabilities end-to-end. Each use case names the goal, walks the flow, and points at the concept and reference pages that hold the deeper detail.
- [AI Intake Pipeline (External App Integration)](https://docs.gloo.com/forge/knowledge-base/use-cases/code-red/ai-intake-pipeline): Use case — an existing Next.js AI application gains governance, token attribution, and centralized agent management by routing inference through Forge as an External App.
- [Dogfood SDLC Pipeline](https://docs.gloo.com/forge/knowledge-base/use-cases/factories/dogfood-sdlc-pipeline): Use case — a live software factory that runs the platform's own SDLC: change requests flow through triage, plan, implement, verify, and release with humans at the gates.
- [Tier 1 Catalog Execution](https://docs.gloo.com/forge/knowledge-base/use-cases/governance/tier1-catalog-execution): Use case — how typed catalog commands flow end-to-end from skill grant to sandbox execution, with every invocation captured in the governance audit log.
- [Tier 3 Per-Command Governance](https://docs.gloo.com/forge/knowledge-base/use-cases/governance/tier3-audit-mode): Use case — every free-form sandbox command flows through a structured policy decision and lands in the audit log with full per-command attribution.
- [Splunk Notable Triage](https://docs.gloo.com/forge/knowledge-base/use-cases/secops/splunk-notable-triage): Use case — automate the first-pass triage of Splunk Enterprise Security notables with a single agent, a structured skill, and per-notable disposition write-back.
- [Org Chart Draft Editor](https://docs.gloo.com/forge/knowledge-base/use-cases/workforce/org-chart-draft-editor): Use case — plan, review, and publish structural changes to the live organization atomically through a staged singleton draft.
- [Wire a Secrets Store Provider — 1Password](https://docs.gloo.com/forge/knowledge-base/wire-secrets-store-1password): Worked example of the capability-provider pattern — replace the internal secrets store with 1Password as the active provider, without touching any call site that resolves credentials.
- [build-xlsx-workbook](https://docs.gloo.com/forge/procedures/build-xlsx-workbook): Builtin Procedure (Python) that builds an Excel workbook from a typed sheets specification. Deterministic, idempotent, formula-injection-safe.
- [filter-by-date-range](https://docs.gloo.com/forge/procedures/filter-by-date-range): Pure-transform Procedure (Python). Reads a JSON array of records, keeps the ones whose date field falls within a closed `[from, to]` interval, writes the filtered array. Deterministic, idempotent, no network egress.
- [http-fetch-with-retry](https://docs.gloo.com/forge/procedures/http-fetch-with-retry): Generic HTTP fetch Procedure with exponential backoff and structured error codes.
- [paginate-linear-issues](https://docs.gloo.com/forge/procedures/paginate-linear-issues): Built-in Python Procedure that exhaustively paginates a Linear workspace across every team and filters issues by a date field within a monthly period. Returns a normalized JSON file.
- [parse-csv](https://docs.gloo.com/forge/procedures/parse-csv): Builtin Procedure that parses a CSV file against a typed schema and returns either inline records or a fileId for large outputs.
- [Quickstart](https://docs.gloo.com/forge/quickstart): Run your first Forge workflow in under ten minutes — define an agent, compose a workflow, watch it execute, see the audit trail.
- [Quickstart — MCP host](https://docs.gloo.com/forge/quickstart/mcp-host): Register Forge as an MCP server in Claude Desktop, Claude Code, or any MCP-compatible host so an agent can call Forge tools natively.
- [Quickstart — Next.js](https://docs.gloo.com/forge/quickstart/nextjs): Wire Forge into a Next.js App Router app — typed SDK in route handlers, server-rendered runs, environment-driven keys.
- [Quickstart — Raw HTTP](https://docs.gloo.com/forge/quickstart/raw-http): Hit the Forge Platform API from any HTTP client — curl, Python requests, Go net/http, anywhere there's bearer auth and JSON.
- [Quickstart — Vercel](https://docs.gloo.com/forge/quickstart/vercel): Deploy a Forge-integrated app to Vercel — env vars, runtime selection, OAuth token caching across cold starts, and the production checklist.
- [API Reference](https://docs.gloo.com/forge/reference): REST-RPC framing for the Forge Platform API — base URL, authentication, response envelope, pagination, rate limiting, the 12-domain surface map, and the three integration paths.
- [Conventions](https://docs.gloo.com/forge/reference/conventions): Envelope shapes, operation naming, path conventions, scope catalog, pagination, and rate limiting for the Forge Platform API.
- [Errors](https://docs.gloo.com/forge/reference/errors): Canonical error codes returned by the Forge Platform API, when each fires, and how to handle them as a consumer.
- [Agents](https://docs.gloo.com/forge/reference/http-api/agents): Author, fetch, and manage agent definitions through the Platform API.
- [Create an agent](https://docs.gloo.com/forge/reference/http-api/agents/create): Creates a new agent definition. Requires scope `agents:write`.
- [Delete an agent](https://docs.gloo.com/forge/reference/http-api/agents/delete): Deletes an agent definition. Requires scope `agents:write`.
- [Get an agent by ID](https://docs.gloo.com/forge/reference/http-api/agents/get): Requires scope `agents:read`.
- [Get the A2A AgentCard for an agent](https://docs.gloo.com/forge/reference/http-api/agents/getCard): Returns the Agent-to-Agent (A2A) discovery card for the given agent. Requires scope `agents:read`.
- [List agents](https://docs.gloo.com/forge/reference/http-api/agents/list): Returns the agent definitions visible to the calling External App. Requires scope `agents:read`.
- [Update an agent](https://docs.gloo.com/forge/reference/http-api/agents/update): Updates an existing agent definition. Requires scope `agents:write`.
- [Budgets](https://docs.gloo.com/forge/reference/http-api/budgets): Read budget status across every scope visible to the caller.
- [Get budget status across all scopes](https://docs.gloo.com/forge/reference/http-api/budgets/getStatus): Returns one entry per budget scope with budget, used, remaining, and percent-used amounts (in cents). Requires scope `cost:read`.
- [Chat](https://docs.gloo.com/forge/reference/http-api/chat): Drive Hybrid Workforce Chat — channels, sessions, and messages — from external systems.
- [Create a chat session in a channel](https://docs.gloo.com/forge/reference/http-api/chat/createSession): Requires scope `chat:write:session`.
- [Get a chat session](https://docs.gloo.com/forge/reference/http-api/chat/getSession): Requires scope `chat:read`.
- [List chat channels](https://docs.gloo.com/forge/reference/http-api/chat/listChannels): Returns channels visible to the calling principal. Requires scope `chat:read`.
- [List messages in a session](https://docs.gloo.com/forge/reference/http-api/chat/listMessages): Requires scope `chat:read`.
- [Send a message to a session](https://docs.gloo.com/forge/reference/http-api/chat/sendMessage): Requires scope `chat:write:message`.
- [Evals](https://docs.gloo.com/forge/reference/http-api/evals): Browse evaluation sets and their results.
- [Get an evaluation set by ID](https://docs.gloo.com/forge/reference/http-api/evals/get): Requires scope `evals:read`.
- [List evaluation sets](https://docs.gloo.com/forge/reference/http-api/evals/list): Requires scope `evals:read`.
- [External App API](https://docs.gloo.com/forge/reference/http-api/external): The /api/v2/external/* surface — endpoints called by external apps to manage their own agents and discover available models. Distinct from the management API and the inference carve-out.
- [Find or create a personal agent](https://docs.gloo.com/forge/reference/http-api/external/findOrCreateAgent): Idempotently provision a PERSONAL agent for a user. Returns the existing agent if one already exists for the (app, user) pair. Requires scope `agents:write`.
- [List available models for an agent](https://docs.gloo.com/forge/reference/http-api/external/listAvailableModelsForAgent): Returns the OpenAI-compatible model list for a specific agent. Resolves by explicit agentId, user JWT, or app default. Requires scope `agents:read`.
- [External Apps](https://docs.gloo.com/forge/reference/http-api/externalApps): Admin-only OAuth2 client management for integrating systems.
- [Create an External Application](https://docs.gloo.com/forge/reference/http-api/externalApps/create): Mints OAuth2 client credentials for a new External App. The `client_secret` is returned ONCE in this response and never persisted in plaintext. Admin-only. Requires scope `externalApps:write`.
- [Deactivate (soft-delete) an External Application](https://docs.gloo.com/forge/reference/http-api/externalApps/deactivate): Marks the App as inactive and archives any agents owned by it. Admin-only. Requires scope `externalApps:write`.
- [Get an External Application by ID](https://docs.gloo.com/forge/reference/http-api/externalApps/get): Admin-only. Requires scope `externalApps:read`.
- [List External Applications](https://docs.gloo.com/forge/reference/http-api/externalApps/list): Admin-only. Requires scope `externalApps:read`.
- [Rotate the client_secret for an External Application](https://docs.gloo.com/forge/reference/http-api/externalApps/rotateSecret): Generates a new client_secret and invalidates the previous one. The new secret is returned ONCE in this response. Admin-only. Requires scope `externalApps:write`.
- [Update an External Application](https://docs.gloo.com/forge/reference/http-api/externalApps/update): Admin-only. Requires scope `externalApps:write`.
- [Factories](https://docs.gloo.com/forge/reference/http-api/factories): Inspect Factories — autonomous production units — and their assigned workers.
- [Get a factory by ID](https://docs.gloo.com/forge/reference/http-api/factories/get): Requires scope `factories:read`.
- [List factories](https://docs.gloo.com/forge/reference/http-api/factories/list): Requires scope `factories:read`.
- [List workers assigned to a factory](https://docs.gloo.com/forge/reference/http-api/factories/listWorkers): Requires scope `factories:read`.
- [Files](https://docs.gloo.com/forge/reference/http-api/files): Upload artifacts and download files produced by workflow runs.
- [Governance](https://docs.gloo.com/forge/reference/http-api/governance): Drive the Trust Fabric approvals queue and inspect active policies.
- [Get an approval by ID](https://docs.gloo.com/forge/reference/http-api/governance/getApproval): Requires scope `governance:read`.
- [List pending approvals](https://docs.gloo.com/forge/reference/http-api/governance/listApprovals): Requires scope `governance:read`.
- [List governance policies](https://docs.gloo.com/forge/reference/http-api/governance/listPolicies): Requires scope `governance:read`.
- [Record a decision on an approval](https://docs.gloo.com/forge/reference/http-api/governance/recordDecision): Approve or reject a pending approval. Requires scope `governance:write:approve`.
- [Knowledge](https://docs.gloo.com/forge/reference/http-api/knowledge): Query and ingest Forge-managed knowledge via vector search and RAG-grounded synthesis.
- [Memory](https://docs.gloo.com/forge/reference/http-api/memory): Read and write agent memory entries across the channel and long-term layers.
- [Create a memory entry](https://docs.gloo.com/forge/reference/http-api/memory/create): Requires scope `memory:write`.
- [Get a memory entry by ID](https://docs.gloo.com/forge/reference/http-api/memory/get): Requires scope `memory:read`.
- [List memory entries](https://docs.gloo.com/forge/reference/http-api/memory/list): Requires scope `memory:read`.
- [Update a memory entry](https://docs.gloo.com/forge/reference/http-api/memory/update): Requires scope `memory:write`.
- [Models](https://docs.gloo.com/forge/reference/http-api/models): Browse the catalog of language models available to agent definitions.
- [Get a model by ID](https://docs.gloo.com/forge/reference/http-api/models/get): Requires scope `models:read`.
- [List model catalog entries](https://docs.gloo.com/forge/reference/http-api/models/list): Requires scope `models:read`.
- [Procedures](https://docs.gloo.com/forge/reference/http-api/procedures): Sandbox-facing HTTP shims that let a running Procedure sandbox proxy capability calls and file IO back through Forge. Authenticated by per-dispatch session token.
- [Get a file](https://docs.gloo.com/forge/reference/http-api/procedures/getFile): Read a file the dispatch has access to from inside a Procedure sandbox. Three modes: text, sandbox, meta. Same schema as the agent-side getFile tool.
- [Invoke a capability](https://docs.gloo.com/forge/reference/http-api/procedures/invokeCapability): Proxy a Convex-side capability provider call from inside a running Procedure sandbox. Authenticated by per-dispatch session token.
- [Put a file](https://docs.gloo.com/forge/reference/http-api/procedures/putFile): Persist file bytes produced inside a Procedure sandbox via the unified files/bindings model. Same schema as the agent-side putFile tool.
- [Runs](https://docs.gloo.com/forge/reference/http-api/runs): Inspect workflow runs and resume runs paused at human or approval gates.
- [Get run detail (with steps)](https://docs.gloo.com/forge/reference/http-api/runs/get): Requires scope `runs:read`.
- [List workflow runs](https://docs.gloo.com/forge/reference/http-api/runs/list): Returns recent runs visible to the calling External App. Requires scope `runs:read`.
- [Resume a paused run](https://docs.gloo.com/forge/reference/http-api/runs/resume): Resumes a run awaiting human input. Requires scope `workflows:write:resume`.
- [Skills](https://docs.gloo.com/forge/reference/http-api/skills): Read and author packaged instruction sets that compose into agent prompts.
- [Create a skill definition](https://docs.gloo.com/forge/reference/http-api/skills/create): Requires scope `skills:write`.
- [Get a skill by ID](https://docs.gloo.com/forge/reference/http-api/skills/get): Requires scope `skills:read`.
- [List skill definitions](https://docs.gloo.com/forge/reference/http-api/skills/list): Requires scope `skills:read`.
- [Update a skill definition](https://docs.gloo.com/forge/reference/http-api/skills/update): Requires scope `skills:write`.
- [Webhooks](https://docs.gloo.com/forge/reference/http-api/webhooks): Subscribe External Apps to HITL and workflow lifecycle events. Forge signs every delivery with the subscription secret and retries on transient failures.
- [Create a webhook](https://docs.gloo.com/forge/reference/http-api/webhooks/create): Create a webhook subscription for an External App. Returns the signing secret once — store it immediately.
- [Delete a webhook](https://docs.gloo.com/forge/reference/http-api/webhooks/delete): Revoke a webhook subscription by id.
- [List webhooks](https://docs.gloo.com/forge/reference/http-api/webhooks/list): List webhook subscriptions owned by the calling External App.
- [Rotate webhook secret](https://docs.gloo.com/forge/reference/http-api/webhooks/rotateSecret): Mint a new signing secret for an existing webhook subscription. Returns the new secret once.
- [Workflows](https://docs.gloo.com/forge/reference/http-api/workflows): Discover workflow definitions, read their input schema, and dispatch new runs.
- [Dispatch a workflow run](https://docs.gloo.com/forge/reference/http-api/workflows/dispatch): Trigger a new run of a workflow by name. Requires scope `workflows:write:dispatch`.
- [Get a workflow definition by name](https://docs.gloo.com/forge/reference/http-api/workflows/get): Resolve a workflow by its kebab-case name identifier. Requires scope `workflows:read`.
- [Get workflow input schema](https://docs.gloo.com/forge/reference/http-api/workflows/getInputSchema): Fetch the active version's input contract for typed UI rendering and pre-dispatch validation.
- [List workflow definitions](https://docs.gloo.com/forge/reference/http-api/workflows/list): Requires scope `workflows:read`.
- [Inference Carve-Out](https://docs.gloo.com/forge/reference/inference): Two OpenAI-compatible surfaces. /api/inference/v1/* is the External-App carve-out (JWKS-signed JWT). /api/llm/v1/* is the LLM proxy (virtual-key auth) backed by the routing tier abstraction.
- [MCP Server](https://docs.gloo.com/forge/reference/mcp-server): Model Context Protocol server that exposes Forge platform capabilities as tools and resources for AI agents.
- [Observability Alerts](https://docs.gloo.com/forge/reference/observability-alerts): Schema, cron semantics, sink interface, and audit footprint for the observeAlertRules table and the evaluateObserveAlerts cron.
- [Procedure SDK (Node)](https://docs.gloo.com/forge/reference/procedure-sdk-node): Reference for @gloo/procedure-sdk-node — the author-facing SDK for Forge procedures written in TypeScript. Every capability call, file read/write, and progress event flows through here.
- [Procedure SDK (Python)](https://docs.gloo.com/forge/reference/procedure-sdk-python): Reference for gloo_procedure_sdk — the author-facing SDK for Forge procedures written in Python. Every capability call, file read/write, and progress event flows through here.
- [Reactive Client](https://docs.gloo.com/forge/reference/reactive-client): The Convex `useQuery` / `useMutation` surface that the Forge Console subscribes to. Same operations as the HTTP API, exposed as live, auto-revalidating subscriptions.
- [Observability](https://docs.gloo.com/forge/reference/reactive-client/observability): Reactive client functions for the observability family — traces, attention items, notifications, dashboards, sandbox lineage, and the cost / token-event surface.
- [Platform](https://docs.gloo.com/forge/reference/reactive-client/platform): Reactive client functions for the platform family — the registry (tools, skills, models, steps), settings (alerts, feature flags, webhooks), and platform-config reads.
- [Production](https://docs.gloo.com/forge/reference/reactive-client/production): Reactive client functions for the production family — workflows, runs, run-step timelines, chat sessions, and dispatch / lifecycle controls.
- [Trust Fabric](https://docs.gloo.com/forge/reference/reactive-client/trust-fabric): Reactive client functions for the Trust Fabric family — approvals, governance policies, audit events, authority grants, and platform kill switches.
- [Workforce](https://docs.gloo.com/forge/reference/reactive-client/workforce): Reactive client functions for the workforce family — agents, workers, factories, teams, and the org-unit tree.
- [Slack Gateway Endpoints](https://docs.gloo.com/forge/reference/slack-gateway): HTTP endpoint reference for the Slack Gateway — Slack-facing routes on the primary deployment, the internal relay surface every deployment exposes, the OAuth callback, action-token format, and the error categories the operator will see in audit.
- [Tool Sets](https://docs.gloo.com/forge/reference/tool-sets): Every code-shipped tool set in the Forge registry — composition, consumers, and notes.
- [TypeScript SDK](https://docs.gloo.com/forge/reference/typescript-sdk): Zero-dependency typed client for the Forge Platform API. Works in Node.js, browsers, and edge runtimes.
- [Run in Production](https://docs.gloo.com/forge/run): Operating Forge in production — Trust Fabric, approvals, budgets, observability, audit, scheduling, kill switches, deployment, factories.
- [Approvals & Human-in-the-Loop](https://docs.gloo.com/forge/run/approvals-and-hitl): Operating the approval surface — how approvals are triggered, the state machine, resolution channels, and the policy rules behind them.
- [Audit & Replay](https://docs.gloo.com/forge/run/audit): The append-only audit trail — what it records, how to query it, replay a run, and the protected-data classifications that gate sensitive content.
- [Budgets & Cost](https://docs.gloo.com/forge/run/budgets-and-cost): Operating envelope and org-unit budgets — scopes, periods, the spend pipeline, and how to read the analytics that fall out of it.
- [Deployment & Webhooks](https://docs.gloo.com/forge/run/deployment-and-webhooks): How agents are activated for dispatch, and how external systems subscribe to Forge events through HTTP webhooks.
- [Factories in Production](https://docs.gloo.com/forge/run/factories): Operating autonomous production units — the Factory Manager, worker types, lifecycle, budget, and how token attribution flows.
- [Judges and Verdicts](https://docs.gloo.com/forge/run/judges-and-verdicts): Operate Judges in production: calibrate against operator overrides, suppress noise per workspace, rate-this-run feedback, drift alerts.
- [Kill Switches](https://docs.gloo.com/forge/run/kill-switches): The operator's stop button — the platform kill switch, agent-level pauses, and how to stop dispatch fast when something is wrong.
- [Observability](https://docs.gloo.com/forge/run/observability): Operating Forge's live and investigative surfaces — what the four signal sources are, where they render, what guarantees they carry, and the rollup, trace-correlation, saved-view, and alert-rule features layered on top.
- [Planning and Goals](https://docs.gloo.com/forge/run/planning-and-goals): Goals, Roadmaps, Workstreams, Milestones, Blockers, Critical Path. The planning surface that ties the workforce's work to outcomes.
- [Scheduling & Triggers](https://docs.gloo.com/forge/run/scheduling-and-triggers): Operating workflow schedules — cron and interval triggers, the polling evaluator, drift prevention, and timezone semantics.
- [Slack in Production](https://docs.gloo.com/forge/run/slack): Standing up and operating the Slack Gateway — install, multi-environment setup, drift detection, and the failure modes an operator should know.
- [Synthesis Budgets and Curator](https://docs.gloo.com/forge/run/synthesis-budgets-and-curator): Operate JIT Capability Synthesis at workspace scale. Monthly cap, alert threshold, Curator review queue, auto-deprecate, drift alerts.
- [Trust Fabric in Production](https://docs.gloo.com/forge/run/trust-fabric): How operators run with the governance pipeline live — coverage, identity, lifecycle, agent trust, and what each disposition means in practice.
- [Webhooks](https://docs.gloo.com/forge/run/webhooks): Operator-facing webhook subscriptions. Encrypted secrets, Splunk HEC format support, idempotencyKey for receiver dedup, replay from the audit log.
- [Error Reference](https://docs.gloo.com/forge/troubleshooting/errors): Every error code an operator may encounter — what triggered it, how to resolve it, and where to look in the audit trail.
- [Welcome to Forge](https://docs.gloo.com/forge/welcome): Forge is the production runtime for AI agents. Define an agent, compose it into a workflow, ship it with approvals, budgets, and audit built in.
- [Why Forge](https://docs.gloo.com/forge/why)