<picture> <source media="(prefers-color-scheme: dark)" srcset="docs/public/logo-dark.svg"> <img src="docs/public/logo-light.svg" width="56" height="56" alt=""> </picture>

Switchboard

Mention it in Slack and an agent reviews the PR, ships the fix, or answers the question — on the model you choose, with its tools running where you decide.

<p align="center"> <img alt="A review in a Slack thread: a plain request to look at a pull request, the card naming the agent it picked and why, then the verdict with its findings." src="docs/public/slack/run.gif" width="395"> </p> <picture> <source media="(prefers-color-scheme: dark)" srcset="docs/public/screenshots/run-page-dark.png"> <img alt="A finished run's page on the dashboard." src="docs/public/screenshots/run-page-light.png" width="100%"> </picture>

Quick start

Node 24 and an Anthropic API key; no Slack.

npx @coreplane/switchboard init --organization <your GitHub org> --anthropic-key <your key>
npx @coreplane/switchboard ask "what can you do?"

init writes .env (mode 600) and config/config.yaml into ~/.switchboard, so every command works from any directory; ask runs the whole pipeline. curl -fsSL https://openswitchboard.dev/install.sh | sh is the same init.

Next: Get started.

How it is put together

Four seams: Channel, Provider, Executor, Agent. Each is an interface with more than one implementation. The dispatcher sits between them: it reads what you asked for and picks the agent, resolves the model through the config layers, checks authorization and runs the agent loop (How a request flows).

<!-- generated:four-seams · npm run docs:gen — drawn from docs/.vitepress/theme/seams.mjs and src/deploy/plan.ts, do not edit by hand -->

flowchart LR
    subgraph channel ["Channel — how a request arrives"]
        C1["Slack"]
        C2["CLI"]
        C3["HTTP · MCP"]
    end
    D{"Dispatcher<br/>routing · config layers · authorization"}
    subgraph agent ["Agent — what runs"]
        AG["general · coding · review · ship · research · explore · conductor"]
    end
    subgraph provider ["Provider — the model"]
        P["Anthropic · OpenAI-compatible"]
    end
    subgraph executor ["Executor — where tools run"]
        E["local · sandbox · resident"]
    end
    C1 & C2 & C3 -->|"message"| D
    D -->|"runs"| AG
    AG <-->|"complete"| P
    AG <-->|"bash · read · write"| E

<!-- /generated:four-seams -->

What you need

| | You need | What it unlocks | |---|---|---| | Required | A model provider key: Anthropic, or any OpenAI-compatible endpoint | Every agent | | For Slack | A Slack app (Socket Mode: bot token + app-level token) | The bot in your workspace | | Optional | A GitHub App, or a repo-scoped token | Repository reads; pull requests | | Optional | A Cloudflare account | Production: the bot plus the state, resident and sandbox Workers | | Optional | An E2B account | Per-thread micro-VMs without Cloudflare | | Optional | A Brave Search key (BRAVE_SEARCH_API_KEY) | Web search for the research agent |

Off means the bot never offers it and the dashboard and the plan never show it (Turn features on and off). The bot is npx @coreplane/switchboard start on any machine with Node, or the published image ghcr.io/coreplanelabs/switchboard.

Learn more

Docs: https://openswitchboard.dev (docs/) · CONTRIBUTING.md · AGENTS.md · SECURITY.md · Apache-2.0.

Going deeper: commands and directives exist for whoever wants to type them (Slack commands); a plain sentence needs none.