anoman
Sign InGet API Key
Documentation home

Agent Integration Guides

Anoman is compatible with any agent framework that supports a configurable base URL. These guides walk through the integration for the most popular frameworks.

Enable streaming to integrate an AI agent

Set stream: true on every request from an autonomous or long-running agent — coding assistants, multi-step chains, agentic loops. Long agent tasks can run past the 100-second edge timeout and fail with a 524 error. Streaming returns the response as it is generated and keeps the connection alive, so the request completes. Most frameworks below enable it with a single flag or setting.

New to streaming? Read the streaming guide.

See every supported agent

The full compatibility matrix — CLI, IDE, framework, and desktop tools, and exactly how each one connects.

Open the compatibility matrix →

Integration guides

  • LangChain · Python — Route LangChain agents through Anoman using the OpenAI-compatible base URL. Works with LangChain Expression Language (LCEL), LangGraph, and any LCEL chain.

  • Claude Code · CLI + Python — Connect Claude Code and any Anthropic SDK client to Anoman using the Anthropic-compatible proxy endpoint. Full guardrail pipeline applied to every request.

  • CrewAI · Python — Configure CrewAI agents to use Anoman as the LLM backend. Route crew tasks through guardrails and policy enforcement with zero code changes to your crew definition.

  • OpenAI Agents SDK · Python — Integrate the OpenAI Agents SDK with Anoman. Swap the base URL on AsyncOpenAI to add guardrails, batch routing, and observability to every agent run.

  • Cursor · IDE — Route Cursor's AI completions through Anoman by setting the OpenAI base URL in Cursor settings. Guardrails apply to every code completion and chat request.

  • Aider · CLI — Use Aider with Anoman by passing the base URL flag. All LLM calls in your aider sessions route through the guardrail pipeline automatically.

  • Continue · IDE Extension — Configure Continue (the open-source AI code assistant) to use Anoman as the model provider. Works with VS Code and JetBrains extensions.

  • AutoGen · Python — Connect Microsoft AutoGen multi-agent workflows to Anoman. Set the base URL on the OpenAI client config to add guardrails to every agent conversation.

  • Cline · VS Code — Configure Cline (the autonomous coding agent VS Code extension) to route requests through Anoman. Set the OpenAI base URL in Cline's provider settings.

  • Zed · Editor — Set an OpenAI-compatible api_url in Zed's settings.json to route the assistant through Anoman for guardrails, routing, and observability.

  • Roo Code · VS Code — Point Roo Code's OpenAI Compatible provider at Anoman so every autonomous edit and command passes through guardrails and tool-call policy.

  • OpenHands · Autonomous Agent — Route OpenHands (OpenDevin) agents through Anoman with a custom LLM base URL. Guardrails and observability on every planning and tool-execution step.

  • LibreChat · Self-hosted Chat — Add Anoman as a custom endpoint in librechat.yaml to give your self-hosted chat UI one gateway to hundreds of LLMs with guardrails on every message.

  • OpenAI Codex CLI · CLI — Connect OpenAI Codex CLI to Anoman with a custom OpenAI-compatible provider in ~/.codex/config.toml. Pin wire_api to "chat" so Codex uses the chat-completions endpoint.

  • OpenClaw · CLI — Add Anoman to OpenClaw as an OpenAI- or Anthropic-compatible provider, matching its openai-completions / anthropic-messages provider config.

  • Goose · CLI — Route Block's Goose agent through Anoman via its OpenAI provider. Set OPENAI_HOST to the host without /v1 — Goose appends the path itself.

  • opencode · CLI — Register Anoman in opencode as an OpenAI-compatible (or Anthropic) provider through the AI SDK — one gateway for every model opencode drives.

  • Hermes · CLI — Point the Hermes terminal agent (Nous Research) at Anoman with a provider using the chat_completions or anthropic_messages transport.

  • GitHub Copilot · VS Code — Send GitHub Copilot chat and agent requests through Anoman with VS Code's Custom Endpoint. Guardrails apply to chat and agent mode; tab completion stays on Copilot's model.

  • Vercel AI SDK · JS/TS — Route the Vercel AI SDK through Anoman with createOpenAICompatible (or createAnthropic). generateText and streamText work unchanged.

  • Pydantic AI · Python — Give Pydantic AI's OpenAI provider a custom base URL to route typed agents through Anoman, with the full guardrail pipeline on every call.

  • Jan · Desktop — Add Anoman to the Jan desktop app as a custom OpenAI- or Anthropic-compatible provider — a local chat UI backed by guardrails.

  • Back to Docs

On this page