What Roam is for
Roam gives coding agents reusable local code analysis and
checks. Its tools return code connections, findings,
candidate algorithmic alternatives, and check results. Agents can use
those results to choose what to inspect, try, or test instead of writing
each analysis from scratch. The local index makes code relationships available to query;
refresh it with roam index as the code changes. Agents use
the tools directly; people set the direction and decide what ships.
Roam can also keep a record of the checks when its run ledger and proof-bundle workflow are enabled. That record helps you review what was checked and which gaps remain. Signed records help reveal later changes to the evidence; a record does not prove who acted or that every relevant check ran. Roam does not certify compliance with any framework.
Beyond code connections, your agent can ask about repeated logic and potentially expensive patterns, such as repeated parsing or lookups inside loops. Algorithm checks can pair a detected pattern with an alternative from a built-in catalog. These findings are leads to investigate, not proof of a bug or a useful optimization. Check the matched code, run tests, and measure the behavior before applying a suggestion.
Roam works alongside your editor, tests, security scanners, and code review. Its local map helps an agent investigate beyond the edited lines. Static analysis can miss connections, and a suggested test list is not test coverage. See where Roam fits.
Why now
A coding agent can produce more code than you can comfortably read line by line. We built Roam to make some of the follow-up questions repeatable: where is the relevant code, what calls it, where is similar logic, and what did the checks actually inspect?
Those questions belong in the agent's workflow, alongside reading source and running tests. A reusable local index lets the agent query relationships without reconstructing them for each question; refresh it when the code changes. Static checks add findings it can investigate without another model call. That is Roam's contribution, not a promise to replace authorization, backups, testing, or deployment controls.
The CLI and MCP server are free under Apache 2.0. Static checks run on your machine without model calls; your coding agent's own model usage and data handling are separate. Paid PR Replay reports add founder review of an agreed scope and a walk-through. Hosted Roam Review and Roam Cloud are planned, not available to subscribe to.
- Try a local sample — the default sample checks the commit range
HEAD~5..HEAD, not five pull-request identities. It needs enough local Git history. - Ask about a paid report — see the existing scopes, prices, credit terms, and exclusions. Agree availability and written terms before work begins.
- Review records for teams — see the evidence checklist, sample records, and their limits.
- Trust posture — data handling, retention, and processing locations per product surface.
Who built it
Roam is built by Dimitris (handle: Cranot), a sole-trader operating from Athens, Greece. Solo founder; the company structure is a Greek atomiki epicheirisi (sole proprietorship). The aim is simple: build something useful and stay independent.
Find me on GitHub, or email [email protected].
What we believe
- Do the repeatable work locally. Build a code map, refresh it when the source changes, and let agents reuse it. Static checks do not need a model call.
- Keep the limits with the result. Missing or unfinished checks are unknown, not a pass. When you enable evidence recording, inspect what it captured and what is absent.
- Keep the engine open. The CLI and MCP server are free under Apache 2.0. Paid report work is separate; using the local tools does not require a purchase.
- Make data handling clear. Ordinary analysis does not automatically upload your source or telemetry. Connected agents may send results to their model provider. Installation, parser downloads, and optional network features are explained on the security page.
- Let people explore without tracking. This site uses no cookies, tracking, or analytics. See how the site handles visits.
What is free, and what is paid?
Start with the free agent tools. If you want help interpreting recent changes, ask about a paid PR Replay report. The report buys review of an agreed scope and a walk-through, not access to locked local checks. Pricing separates these options from proposed hosted products; no hosted launch date is confirmed.
Roadmap
We want agents to spend less effort reconstructing code connections and leave a clearer record of what they checked. Improvements should earn their place on real repository tasks. See the changelog for released changes and unreleased work; a roadmap or an unreleased entry is not an available product.
Where to find us
- Site: roam-code.com
- Docs: roam-code.com/docs/
- GitHub (engine, Apache 2.0): github.com/Cranot/roam-code
- PyPI (free CLI): pypi.org/project/roam-code
- Paid PR Replay audit: /audit
- Governance evidence pack: /governance · trust posture: /trust
- Pricing: /pricing · compare with semantic reviewers: /compare
- Email: [email protected]
- Security disclosures: [email protected] (see policy)
Want to write about Roam? See the press kit for logos and screenshots.