# Coder
**The runtime for agents, your rules, any model, everywhere.**
Coder turns any model into an agent you own. One runtime and one set of rules, with full control of every task, whether it runs in your local harness, from your own code, or always on in Coder Cloud.
## Why Coder
- **Any harness, any model.** Claude Code, Codex or any other harness, running any model from your subscriptions, an API endpoint, or your own machine. Pick per task.
- **Clean context.** Your conversation stays about intent. The work happens in subagents, and only the result comes back.
- **Fast dispatch.** Handoffs are instant, and light enough to run many agents at once.
- **Fully controlled.** One permission model across every engine, live visibility into every task, and steering mid-run.
- **Always on.** Your agents run in the cloud, connected to GitHub, Slack and any other app, each task in its own workspace with only the access you grant.
- **Your agents, your team.** Build agents with their own instructions, triggers and tools, and share them, with their credentials, across your workspace.
- **Flows and SDK.** Orchestrate whole waves of tasks from one TypeScript file, or drive everything from your own code.
- **Built-in review.** Review any diff with a read-only agent that checks every finding against your repo.
## Get started
| I want to | Start with |
| ---------------------------------------------------------------------------------- | ----------------------------------------------- |
| Hand tasks off from Claude Code, Codex or another harness | [In your local harness](#in-your-local-harness) |
| Run my agents always on in the cloud, connected to GitHub, Slack and any other app | [Coder Cloud](#coder-cloud) |
| Script tasks, reviews and flows from code | [From code](#from-code) |
| Run it on my own servers | [Self-host](#self-host) |
## In your local harness
**Install.**
```bash
npm install -g @wular/coder
```
**Add it to your host.**
| Host | Install |
| -------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Claude Code | `coder setup-host claude`
or, inside Claude Code:
`/plugin marketplace add muzam1l/coder`
`/plugin install coder@coder-plugins`
`/coder:setup` |
| Codex, Pi, OpenCode and any host that reads `~/.agents/skills` | `coder setup-host agents` |
**Connect a model.** Coder uses the engines you're signed in to, or your own model ([Models](docs/models.md)).
```bash
npm install -g @openai/codex && codex login
# and/or
npm install -g @anthropic-ai/claude-code && claude auth login
```
**Use it.** Ask your host:
> Use Coder to explain the directory structure of the workspace.
To make it the default, add this line to your project's `AGENTS.md` or `CLAUDE.md`:
```md
Always use Coder for implementation and exploration.
```
Every task runs under the permissions you set and stays yours to watch, steer or stop. See them all in the browser:
```bash
coder dash
```
> **Recommended setup:** Claude Code as the host (Fable, low or medium effort) and Codex (sol) as the engine, the best split of performance and cost.
[Models](docs/models.md) · [Configuration](docs/config.md) · [MCP](docs/mcp.md)
## Coder Cloud
Your agents, always on at **[coder.wular.ai](https://coder.wular.ai)** or [on your own servers](#self-host). Connect GitHub, Slack or any other app, and they work right inside it.
- **Your agents.** Instructions, triggers and tools, set in the dashboard or in `.coder/agents/` in your repo.
- **Every app, one dashboard.** GitHub, Slack and the rest of the integration catalog.
- **Scoped access.** Each task gets a fresh workspace and a platform token limited to that task.
- **A built-in agent.** `coder` answers questions, reviews pull requests, and opens them for requested changes.
[Dashboard](docs/dash.md) · [Agents](docs/agents/index.md)
## From code
Your host drives tasks itself, but everything is scriptable:
```bash
coder run "" # dispatch a task (--wait to block for the answer)
coder list # recent tasks (running + just stopped)
coder result [task-id] # status + final answer (--wait blocks until done)
coder task steer "" # continue a task with new instructions
coder task ask "" # ask about a task without interrupting it
coder task stop # interrupt it
coder task watch [task-id] # stream the live transcript
coder review --pr 42 --post # read-only review, posted inline
```
Any `[task-id]` defaults to the most recent task.
For a repeatable workflow, ask your host to write a flow:
> Create a Coder flow that fixes every failing test file: one task per file, gated on its tests passing.
A flow is one TypeScript file that fans out tasks, with verification gates and resume after a crash:
```ts
// .coder/flows/fix-tests.ts
import { z } from 'zod';
import { task, gate, pipeline } from '@wular/coder/flow';
const failing = await task('Run `bun test` and list the failing test files.', {
returns: z.object({ files: z.array(z.string()) }),
});
export default await pipeline(
failing.data.files,
file => task(`Fix the failing tests in ${file}.`, { name: file }),
(r, file) => gate(`bun test ${file}`),
);
```
[CLI](docs/cli.md) · [Flows](docs/flows.md) · [Review](docs/review.md) · [SDK](docs/sdk.md)
## Self-host
Run all of Coder Cloud on your own infrastructure: one server, one Postgres database, and your tasks running wherever you choose, from Docker and Vercel Sandbox to GitHub Actions or a teammate's machine. Deploy it in one click on Vercel or anywhere Bun runs.
[Self-hosting](docs/self-host.md) · [Vercel template](src/server/deploy/vercel/README.md)
## Docs
Everything else is at **[coder.wular.ai/docs](https://coder.wular.ai/docs)**.
`coder --help` lists every command, and `coder docs [topic]` prints the bundled guides. `coder upgrade` keeps the CLI and host plugins current.