**OpenABCode** is an LLM-routing coding agent that dynamically routes development tasks to the best-suited models:
- Google ecosystem development tasks are routed to Gemini.
- Code review and testing are routed to ChatGPT.
- All other development tasks are routed to Claude.
To learn more about OpenABCode:
* [Visit openabcode.com](https://openabcode.com), the project website
* [Read the documentation](https://openabcode.com/docs)
## Quick Start
Install with the hosted installer:
```bash
curl -fsSL https://openabcode.com/install.sh | sh
```
Or use a package manager:
```bash
npm install -g @openabcode/coding-agent --ignore-scripts
bun install -g @openabcode/coding-agent
brew install matrixmapai/tap/openabcode
```
Then start OpenABCode:
```bash
openabcode
```
To set up Route mode:
1. Run `/login` and configure the providers you want to use.
2. Run `/route-model` and select the authenticated model that classifies tasks. This does not change the active execution model.
3. Run `/model` and pick one specific model per family — e.g. `gpt-6-astra` for OpenAI, `gemini-3.8-flash` for Google, `claude-fable-5` for Anthropic. Route always executes tasks with the exact model you picked for the chosen family.
4. Run `/route` and select `on`.
When Route is on, the footer shows the configured execution models. Route works as long as at least two model families are configured — with a family missing, routing degrades to the available ones. Every completed routing decision is stored in the session JSONL for audit, and manually overriding a routed model is recorded as veto feedback.
## Customizing Route Rules
Route rules, heuristic keywords, file extension mappings, project markers, and the default provider can all be customized in `~/.openabcode/agent/settings.json` (global) or `.openabcode/settings.json` (project-level). When a field is set, it fully replaces the corresponding built-in default.
```json
{
"router": {
"rules": {
"openai": "Test and automation — algorithms, code review, testing, data analysis, scripting",
"google": "Google ecosystem — Android, Flutter, Firebase, Google Cloud",
"anthropic": "General development — all code writing, editing, debugging, architecture, UI"
},
"keywords": {
"openai": ["algorithm", "unit test", "benchmark", "data analysis", "pipeline"],
"google": ["android", "flutter", "dart", "firebase", "gcp"],
"anthropic": ["refactor", "debug", "architecture", "migration"]
},
"fileExtensions": {
".dart": "google",
".rs": "anthropic",
".ts": "anthropic",
".sh": "openai"
},
"projectMarkers": {
"pubspec.yaml": "google",
"cargo.toml": "anthropic",
"package.json": "anthropic",
"main.tf": "openai"
},
"defaultProvider": "anthropic"
}
}
```
Each field is optional and independent — only configure what you want to override.
## Architecture
```text
User prompt
-> @openabcode/coding-agent Route configuration and model selection
-> @openabcode/ai Classifier and provider API requests
-> @openabcode/agent-core Agent loop, state, and tool execution
```
## All Packages
| Package | Description |
|---------|-------------|
| **[@openabcode/coding-agent](packages/coding-agent)** | Route-first terminal coding agent CLI |
| **[@openabcode/tui](packages/tui)** | Terminal UI renderer behind OpenABCode's interactive editor, model selectors, Route footer, and session views |
| **[@openabcode/ai](packages/ai)** | Unified multi-provider LLM API (OpenAI, Anthropic, Google, etc.) |
## Permissions & Containerization
OpenABCode does not include a built-in permission system for restricting filesystem, process, network, or credential access. By default, it runs with the permissions of the user and process that launched it.
If you need stronger boundaries, containerize or sandbox OpenABCode. See [packages/coding-agent/docs/containerization.md](packages/coding-agent/docs/containerization.md) for three patterns:
- **Gondolin extension**: keep `openabcode` and provider auth on the host while routing built-in tools and `!` commands into a local Linux micro-VM.
- **Plain Docker**: run the whole `openabcode` process in a local container for simple isolation.
- **OpenShell**: run the whole `openabcode` process in a policy-controlled sandbox.
## Contributing
See [CONTRIBUTING.md](CONTRIBUTING.md) for contribution guidelines and [AGENTS.md](AGENTS.md) for project-specific rules (for both humans and agents).
## Development
```bash
npm install --ignore-scripts # Install all dependencies without running lifecycle scripts
npm run build # Build all packages
npm run check # Lint, format, and type check
./test.sh # Run tests (skips LLM-dependent tests without API keys)
./openabcode-test.sh # Run openabcode from sources (can be run from any directory)
```
## Supply-chain hardening
We treat npm dependency changes as reviewed code changes.
- Direct external dependencies are pinned to exact versions. Internal workspace packages remain version-ranged.
- `.npmrc` sets `save-exact=true` and `min-release-age=2` to avoid same-day dependency releases during npm resolution.
- `package-lock.json` is the dependency ground truth. Pre-commit blocks accidental lockfile commits unless `OPENABCODE_ALLOW_LOCKFILE_CHANGE=1` is set.
- `npm run check` verifies pinned direct deps, native TypeScript import compatibility, and the generated coding-agent shrinkwrap.
- The published CLI package includes `packages/coding-agent/npm-shrinkwrap.json`, generated from the root lockfile, to pin transitive deps for npm users.
- Release smoke tests use `npm run release:local` to build, pack, and create isolated npm and Bun installs outside the repo before tagging a release.
- Local release installs, documented npm installs, and `openabcode update --self` use `--ignore-scripts` where supported.
- CI installs with `npm ci --ignore-scripts`, and a scheduled GitHub workflow runs `npm audit --omit=dev` plus `npm audit signatures --omit=dev`.
- Shrinkwrap generation has an explicit allowlist for dependency lifecycle scripts; new lifecycle-script deps fail checks until reviewed.
## License
AGPL-3.0
Portions derived from [Pi](https://github.com/earendil-works/pi) remain licensed under the MIT License. See [NOTICE](NOTICE) for attribution.