Documentation map
View SourceStart with a task guide, then follow its API links for exact options and return values. These pages describe ExAgent 2.x; use Migration when upgrading an application from 1.x.
Learn and build
| Your task | Guide |
|---|---|
| Install and run your first agent | Getting started |
| Call application code and return an Ecto struct | Tools and output |
| Give the model task instructions it loads on demand | Skills |
| Select a model, provider route and budget | Models and limits |
| Keep conversation history and stream UI updates | Runtime and events |
| Persist state, approve a tool and resume work | Durability and approvals |
| Coordinate participants, delegate or run a workflow | Coordination |
| Write deterministic tests | Testing |
| Export traces and metrics | Observability |
| Connect external tools | MCP |
| Diagnose a failed operation | Troubleshooting |
Integration recipes
| Application need | Recipe |
|---|---|
| Persisted approval and queue wake-ups | Continuation jobs |
| Phoenix, LiveView and Oban | Framework integrations |
| Durable sequences and selected specialists | Coordination recipes |
| External data supplied through a tool | Retrieval |
| Resuming an approved MCP request | MCP continuation binding |
| OTLP through a disposable exporter VM | Isolated exporter |
| OTLP through an application-owned Collector | Collector route |
Find a contract
- Architecture explains ownership between layers.
- Supported features and limits lists profiles and deployment requirements.
- Known limits explains the boundaries that affect production handling.
- Migration gives the steps for upgrading an existing application.
- Coding-agent integration notes map tasks to public APIs and result shapes.
- The Modules tab contains the complete public module and function reference.
The same site provides llms.txt, per-page Markdown and an EPUB. Use the Markdown
pages and API links when integrating ExAgent with a coding agent.