LLM tools for your existing app.
appctl reads your API, database, or framework project, generates a local tool contract, and executes the actions your configured model requests. Your app stays the source of truth; appctl handles auth, safety, and audit history.
What appctl does
It is a CLI and local control plane that connects an LLM to the operations your app already exposes. Use it to inspect data, run API actions, operate internal tools, or serve a web console without writing a custom agent bridge.
- Sync from OpenAPI, SQL, Supabase, MCP, or supported frameworks.
- Write `.appctl/schema.json` and `.appctl/tools.json` beside your project.
- Run `appctl chat`, `appctl run`, or `appctl serve` to execute model tool calls.
- Keep secrets in config or environment variables, not in the chat transcript.
Use your favorite AI to set up appctl
Copy the prompt below into ChatGPT, Claude, Gemini, or any agent. It asks them to guide install, sync, and first chat — step by step.
You're my setup assistant. I want to use appctl — it connects real backends (OpenAPI, databases, frameworks) to LLM agents as callable tools. Goal: get me from zero to a working flow where I can chat against my synced tools safely. Do this: 1. Ask what OS I'm on and whether I prefer `cargo install` or a release binary. 2. Give exact commands to install appctl and verify it (`appctl --help`). 3. Walk me through an app directory with `.appctl`, configuring a model provider without pasting secrets into chat, and syncing my schema (I'll describe my stack next). 4. Explain when to use `appctl chat`, `appctl run`, and `appctl serve`; start read-only and conservative confirmations. 5. End with one minimal first prompt I can try after sync. Rules: one question at a time when you need info; short steps; concrete commands; don't invent URLs or credentials — ask me. Official docs: /appctl/docs/introduction/ Repository: https://github.com/Esubaalew/appctl
Why teams use it
Your app stays yours
appctl reads your routes or schema and calls your existing endpoints. No data export, no rewrite, no new database.
Any LLM works
Bring OpenAI, Anthropic, Google, Azure OpenAI, Ollama, LM Studio, vLLM, or any OpenAI-compatible endpoint. Switch per project.
Every call is audited
Each tool call is logged with provenance, arguments, and status in SQLite. Replay runs or export to your own sink.
Start with one synced app.
Run the guided setup, inspect the generated tools, then try a read-only prompt first.