Use cases

Start on one laptop.
Grow to the whole company.

The serverless context layer for AI agents over your company’s knowledge. Everyone sees only what they may.

Just you

A folder on your laptop.

  • Agent memory

    Your agent searches your notes, tickets and chats, so it remembers what you already decided.
    • Local files
    • Slack
    • Jira
    • Notion
  • Coding context

    Before the first line, your coding agent reads the ticket, the pull requests and the threads behind them.
    • GitHub
    • Jira
    • Slack
  • Local docs

    Search folders of Markdown, text and code by meaning and by exact words, on your own disk.
    • Local files

Your team

One S3 bucket. No server.

  • Shared knowledge

    Index the team’s docs, chats, tickets and code once. Every teammate’s agent searches the same index.
    • Google Drive
    • Notion
    • Slack
    • GitHub
  • Code search

    Search code, pull requests and issues across your repositories, by meaning and by exact names.
    • GitHub
    • Local files
  • Support

    Answer customers from product docs, past tickets and the threads where the fix was explained.
    • Jira
    • Slack
    • Notion
    • Google Drive

The company

One reader behind your login.

  • Company search

    One search across every source, from the AI apps people already use, behind your company login.
    • Any MCP client
    • OIDC login
  • Permissions kept

    fyrr checks each person’s access inside the search, so nobody sees a document they could not open.
    • Keycloak
    • SCIM
  • Your cloud

    The index and its permissions live in your own S3 bucket. No database or search cluster to run.
    • S3 bucket
    • Local disk

Builders

Your own agents, models and sources.

  • Agent tools

    Give the agents you build search, fetch and filters over MCP, for one user or many.
    • MCP stdio
    • MCP HTTP
  • Train & evaluate

    Train and test search agents on one fixed snapshot of the index, so every run sees the same documents.
    • Snapshots
    • Rollback
  • Any source

    Write a connector for any internal system. Documents and permissions sync like built-in ones.
    • Python SDK

Get started

Chart your
own islands.

$ git clone https://github.com/fyrr-ai/fyrr && cd fyrr
$ scripts/doctor --local
$ examples/quickstart/run

Pre-release: builds from source on Linux; the build prerequisites (including openssl and jq) are listed in CONTRIBUTING. Cloud connectors are tested against synthetic APIs so far.