Guides

Practical how-to documentation for fullsend, organized by audience. For design documents and architectural context, see docs/problems/, docs/ADRs/, and docs/architecture.md.

Getting started

Guides for onboarding organizations and configuring GitHub — the first thing most users need.

Mint enrollment — Enroll your org or repo in a token mint before configuring anything else
Getting Inference — Provision GCP inference access for your org or repo
Configuring GitHub — Install GitHub Apps and run the setup CLI
Organization Mode — (deprecated — see per-repo Getting Started) Org-wide setup with a shared .fullsend config repo

Guides for organization owners and repository administrators who manage fullsend installations.

Operations — Enrollment, configuration updates, status checks, uninstall, and standalone commands
Advanced setup — Alternative installation paths, setup flags, custom app sets, and manual WIF configuration
OpenAI Workload Identity — Run GPT models on the pi runtime without a stored OpenAI key: console setup, repository variables, local runs, troubleshooting

Advanced guides for platform operators who deploy and manage the GCP-side infrastructure (token mint, WIF, secrets).

Mint service administration — Deploying and managing the token mint (GCP or Cloudflare)
Standalone mint — Running the token mint as a standalone HTTP server without GCP
Infrastructure reference — Token mint, WIF, and secrets deployment details
Enabling fullsend on private repositories — Additional guardrails and configuration for private repos
Tracing reference — Telemetry levels, environment variables, span hierarchy, and attributes
Eval measurements — Online trace scoring with eval-measurements.jsonl and measurement manifests

Guides for developers working in repositories where fullsend is active.

Bugfix workflow — End-to-end guide to how fullsend handles a bug report from issue to merge
Issue commands — Slash commands and label triggers for interacting with agents
Running agents locally — Run fullsend agents on your machine using released binaries (macOS + Linux)

Start with the overview to pick the right approach, then follow the relevant guide:

Customizing agents overview — Quick decision guide for all customization approaches
Configuring with AGENTS.md — Guide agents using your repo's AGENTS.md file
Configuring with skills — Extend built-in agent skills; authoring augmentations
Configuring agent behavior — Harness composition, status notifications, and disabling agents
Bring Your Own Agent — Build and register a custom agent from scratch
Custom Agent Identity — Using a standalone mint for custom GitHub App identity
Harness Field Reference — Complete harness YAML field reference, merge rules, and resource referencing
CEL Triggers Reference — Dispatch flow, NormalizedEvent fields, transition kinds, and trigger patterns
Custom Poller Example — Create a custom poller workflow that invokes fullsend harness agents with a pre-computed matrix
Building custom agents from scratch — (deprecated — see Bring Your Own Agent)
Default, derived, and custom agents — When configuration crosses into derived or custom agent territory

Jira Integration — Connect fullsend to a Jira project so that issue comments and label changes trigger agents
How to emit traces — Configure a repository or organization to send OpenTelemetry traces to a remote backend
Tracing with MLflow — MLflow-specific setup: experiment routing, Basic auth encoding, org-level organization, and cost column caveats

Guides for contributors developing and testing fullsend itself.

E2E testing — Local and CI e2e runs, including PR authorization and ok-to-test
CLI internals — Command structure, installation pipeline, and sandbox runtime
Behaviour testing — Write Gherkin scenarios for end-to-end agent behaviour
Behaviour test drivers — Implement SCM and CI drivers for behaviour tests
Testing workflow changes — Point a live GitHub org at a branch to test workflow, action, and agent changes before release
Tracing internals — How the distributed tracing implementation works and how to extend it