What is Sinatra?

Sinatra is an autonomous coding agent that lives inside Linear and GitHub.

The flow

1. A teammate creates a Linear issue and either assigns it to @sinatra or @-mentions the agent in a comment.
2. The Sinatra API receives a webhook from Linear, looks up the right workspace, and starts a Temporal workflow.
3. A worker spins up an isolated sandbox, clones the target repo, and runs an agent (Claude Code, OpenCode, etc.) with the ticket description as context.
4. The agent investigates, writes code, runs tests, and commits.
5. Sinatra opens a GitHub pull request and posts back to the Linear ticket with a link.

The whole loop is built on durable workflows — if anything crashes, it picks up where it left off rather than restarting from zero.

Why it exists

Most coding tasks teams ship are not deeply novel — they're well-scoped tickets where the hard part is getting started. Sinatra is built for those tickets, the ones where the user can describe what they want in a few sentences and review a PR is faster than writing the code by hand.

It's not trying to replace senior engineering judgment. It's trying to clear out the queue.

What you get

We run the platform — API, worker, Postgres, and Temporal. You connect Linear and GitHub from the dashboard, bring your own LLM key (or use ours), and assign issues. There's nothing to deploy or operate.

What you need

• A GitHub organization or account where you can install the Sinatra GitHub App.
• Optionally, a Linear workspace to connect — or run on GitHub issues alone.
• Model access: an LLM provider key for Anthropic or OpenRouter, a connected Codex / ChatGPT subscription (this is how you reach OpenAI models), or Sinatra's managed key. The agent's intelligence comes from these models; Sinatra is the orchestration layer around them.