# Plico

> Plico is a TypeScript kit that turns a folder of Markdown instructions, skills and TypeScript tools into a runnable, inspectable AI agent: runs, an event log, human approvals, evals and a local Studio. Developer preview, published on npm as `@plico-ai/*` and `create-plico`. Needs Node 22.18 or later.

Start with `npm create plico@latest my-agent`, then read Project layout, Tools, Approvals and Evals. A project is `agent.md`, `plico.config.ts`, `skills/`, `tools/`, `evals/`, `artifacts/` and `memory/`. Verify work with `plico validate`, `plico eval` and `plico run --dry`, none of which need an API key. Every page below is plain Markdown. Full docs in one file: https://plico.sohab.dev/llms-full.txt

- [Welcome to Plico](https://plico.sohab.dev/llms.mdx/docs/content.md): A folder of Markdown and TypeScript becomes a runnable, inspectable agent app.
- [Quickstart](https://plico.sohab.dev/llms.mdx/docs/quickstart/content.md): Scaffold a project, validate it, run it with a model, and approve a tool call in Studio.

- **Build**
- [Project layout](https://plico.sohab.dev/llms.mdx/docs/project-layout/content.md): The files and folders of a Plico project.
- [Tools](https://plico.sohab.dev/llms.mdx/docs/tools/content.md): Write typed tools in TypeScript with defineTool.
- [Approvals](https://plico.sohab.dev/llms.mdx/docs/approvals/content.md): Hold a tool call until a person approves it.
- [Evals](https://plico.sohab.dev/llms.mdx/docs/evals/content.md): Scripted test cases that run with no model and no API key.
- [Model providers](https://plico.sohab.dev/llms.mdx/docs/providers/content.md): Choose a model provider, keep keys in .env, and reach any other provider.
- [Artifacts](https://plico.sohab.dev/llms.mdx/docs/artifacts/content.md): The versioned workspace a run reads and writes.
- [Example: a refund agent](https://plico.sohab.dev/llms.mdx/docs/refund-agent/content.md): Two tools, one that waits for approval, and three evals. Copy it and run it with no key.
- [Build with an AI coding agent](https://plico.sohab.dev/llms.mdx/docs/agents/content.md): Point Claude Code, Cursor or any coding agent at Plico and get a working agent project.

- **Run**
- [Studio](https://plico.sohab.dev/llms.mdx/docs/studio/content.md): Inspect runs, approve tool calls, and chat with your agent.
- [Threads](https://plico.sohab.dev/llms.mdx/docs/threads/content.md): Multi-turn conversations stored alongside runs.
- [Use Plico from your backend](https://plico.sohab.dev/llms.mdx/docs/backend/content.md): Run an agent inside your own server, one user at a time.
- [HTTP API](https://plico.sohab.dev/llms.mdx/docs/http-api/content.md): The native HTTP API served by plico serve.
- [CLI](https://plico.sohab.dev/llms.mdx/docs/cli/content.md): The plico commands.
- [Events](https://plico.sohab.dev/llms.mdx/docs/events/content.md): The event log every run writes, and how to read it.
- [OpenTelemetry tracing](https://plico.sohab.dev/llms.mdx/docs/tracing/content.md): The spans Plico emits for every run, and how to export them.
- [Upgrading](https://plico.sohab.dev/llms.mdx/docs/upgrading/content.md): What a Plico upgrade changes, and how to handle runs and saved data.
- [Troubleshooting](https://plico.sohab.dev/llms.mdx/docs/troubleshooting/content.md): Fix the errors a first project is most likely to hit.

- **More**
- [How Plico compares](https://plico.sohab.dev/llms.mdx/docs/compare/content.md): Where Plico fits among TypeScript agent tools, and what it does not do yet.