Copy this file to
AGENTS.local.mdand fill in your details.AGENTS.local.mdis gitignored (*.local) — it will never be committed. It is read by AI agents afterAGENTS.mdto enrich context with your local setup.
| Variable | Value |
|---|---|
ROUTERLY_HOME |
~/.routerly/ |
| Service port | 3000 (default) |
| Node version | v22.x |
| OS | macOS |
| Key | Value |
|---|---|
| Routerly username | YOUR_USERNAME |
| Test project token | YOUR_TEST_PROJECT_TOKEN |
The test project token is used for all manual API tests and integration tests.
List the LLM providers you have running/configured locally. Agents use this to avoid suggesting configurations that don't match your actual setup.
| Provider | Endpoint | Notes |
|---|---|---|
| Ollama | http://localhost:11434 |
models: llama3, mistral |
| OpenAI | cloud | key in ~/.routerly/config.json |
| Anthropic | cloud | key in ~/.routerly/config.json |
Document branches, features in progress, or pending tasks that agents should know about so they don't create conflicting changes.
# Example:
# branch: feat/streaming-v2
# WIP: refactoring provider adapter interface in packages/service/src/adapters/
# do not touch: packages/service/src/adapters/openai.ts — mid-refactor
Override or extend project-wide coding preferences for this developer.
# Examples:
# - Prefer explicit return types on all functions
# - When in doubt, ask before modifying test files
# - Console.log is acceptable during local development
# - Preferred test runner flags: --reporter=verbose
Tools installed on this machine relevant to the project.
# Example:
# - Docker Desktop 4.x
# - Postman / httpie for API testing
# - pino-pretty installed globally: npm i -g pino-pretty
# - jq available in PATH
Any free-form context the agent should keep in mind when working on this machine.
# Example:
# - This machine is used for both CLI and service work
# - dashboard hot-reload on port 5173 is always running during development
# - Do not run `npm run build` on the full monorepo — it takes >2 min