Snipet

Open source and self-hosted. Cloud is coming soon.

Build AI agents that work across all your apps.

Connect agents to the tools you already use and call them from anywhere: your product, your backend, and soon the channels your users live in.

Get startedStar on GitHub
Ops agentWorking

From zero to a working agent

Four steps in the app, no code until you embed it.

  1. 1

    Connect a model

    Add a provider: OpenAI, Gemini, Mistral, Groq, OpenRouter, or run models locally with Ollama and LM Studio.

    gpt-4.1 via OpenRouter
  2. 2

    Enable plugins

    Turn on integrations from the catalog or point to any MCP server. Credentials are stored encrypted.

    GitHub, Linear
  3. 3

    Create an agent

    Give it a prompt, pick its model and choose which plugins it may use.

    Ops agent
  4. 4

    Use it anywhere

    Call it through the API or drop the chat widget into your app. External channels are next.

    Widget on your site

Any model, no lock-in

Cloud providers or models running on your own hardware. Switch without rewriting the agent.

Plugins your agents can use

Start from the catalog or connect any MCP server. More integrations are on the roadmap.

  • Apify

    Run Apify Actors to scrape and automate thousands of websites

  • AWS Knowledge

    Search AWS documentation, API references and best practices

  • Cloudflare Docs

    Search Cloudflare's developer documentation

  • Context7

    Up-to-date documentation and code examples for libraries and frameworks

  • DeepWiki

    Ask questions about public GitHub repositories

  • Exa

    Web search and code context search powered by Exa

  • Firecrawl

    Scrape, crawl and search websites, returning clean markdown

  • GitHub

    Manage repositories, issues and pull requests through GitHub's hosted MCP server

  • GitMCP

    Documentation of any public GitHub repository (pass owner/repo to the tools)

  • Hugging Face

    Search models, datasets, Spaces and papers on the Hugging Face Hub

  • Linear

    Manage Linear issues, projects and comments

  • Microsoft Learn

    Search and fetch official Microsoft and Azure documentation

  • Neon

    Manage Neon Postgres projects, branches and run SQL

  • Stripe

    Manage Stripe customers, payments, invoices and products

  • Supabase

    Manage Supabase projects, databases, edge functions and docs

  • Tavily

    Web search, extraction and crawling optimized for AI agents

  • +

    Any MCP server

    Bring your own server by URL.

Call your agents from anywhere

Today: the API and an embeddable chat widget for any site, React or Angular. The widget is just the first channel.

Available now

  • API
  • Chat widget

Coming next

  • WhatsApp
  • Telegram
  • Slack
  • Discord
<script src="https://unpkg.com/@snipet/widget/dist/widget.js"></script>
<script>
  SnipetWidget.init({
    apiUrl: "https://snipet.example.com",
    appId: "<your app id>",
    agentId: "<your agent id>",
  });
</script>

Roadmap

What shipped recently and what is coming next.

  1. 0.2.0

    Shipped · Oct 22, 2026

    - Add debug mode

    - Add settings

    - Add anonymous login to widget

    - Add libraries to use widget with web frameworks

  2. 0.3.0

    Shipped · Oct 23, 2026

    - Improve plugin with instructions

    - Fix CORS issue on emit token

    - Create a landing page with documentation

    - Add full screen size for widget

  3. 0.4.0

    Planned · Oct 30, 2026

    - add commands to agent

    - add support for mcp server prompts

    - add support for internal builtin commands

    - add snipet owner command pattern

    - add support for claude skills and other ai agents

Run it on your servers

Snipet is open source and needs only Postgres. Your keys, your data, your infrastructure.

Terminal
curl -O https://raw.githubusercontent.com/usesnipet/snipet/main/docker-compose.yml
docker compose up -d

Then open localhost:8852. Prefer to build from source? Follow the setup guide.

Or let us host it

Snipet Cloud is coming. Join the waitlist and we'll email you when it opens.

Frequently asked questions

What is Snipet?

Snipet is an open-source platform to build AI agents that use your tools, and to put them in front of your users: in a chat widget on your site or through the API from your own product.

Is Snipet free?

Yes. Snipet is open source and you can self-host it on your own servers at no cost. A hosted version, Snipet Cloud, is coming soon; join the waitlist to hear when it opens.

Which LLM providers does Snipet support?

Cloud providers such as OpenAI, Google Gemini, Mistral AI, Groq, DeepSeek, xAI, Cerebras, Fireworks AI, OpenRouter and any OpenAI-compatible API, plus models running on your own hardware with Ollama or LM Studio. You can switch an agent's model without rewriting it.

Can Snipet agents use MCP servers?

Yes. Pick a plugin from the catalog (GitHub, Linear, Stripe, Supabase, Neon and more) or connect any MCP server by its URL. Snipet fetches the server's actions, keeps them in sync and stores the credentials encrypted.

How do I add an AI agent to my website?

Create an app in Snipet, then add the chat widget to your site with a script tag, the npm package, or the React and Angular libraries. To build your own interface instead, chat with the agent through the API.

How do I self-host Snipet?

Snipet is one server plus a Postgres database. Download the docker-compose.yml from the repository, run docker compose up -d and open localhost:8852. You can also run it from source with Node.js 22 and pnpm.

Do my users need an account to chat with an agent?

No. An app can let visitors chat anonymously, or your backend can issue a token that signs your existing users in to the widget.