Roadmap¶
WhyGraph works today for the core loop - scan a repo, serve evidence and rationale over MCP - through a local portal that holds all your projects, with an Explorer and a chat assistant over the same data. Here's what shipped recently, and what's planned but not yet built. Treat the planned items as direction, not a promise of dates.
Recently shipped¶
| Feature | What it gives you |
|---|---|
| The portal | One local server for all your projects: whygraph up, add repos from the browser, an HTTP MCP endpoint per project, scans run and tracked for you. |
| Explorer | A web view over the graph, evidence, and rationale. |
| Chat assistant | Ask questions in English; it calls WhyGraph's tools, runs aggregate SQL, and charts the results. |
| Author identity | One row per human, resolved from mailmap and GitHub rather than guessed. |
| Branch membership | Shipped history is distinguished from work in progress, recomputed and self-healing every scan. |
| Auto-rescan git hooks | The portal tracks your commits in the background; hooks just ask it to rescan. |
| Projects from a platform | Link a local checkout to a project on a production WhyGraph: your agent gets the team's history through your local portal while your uncommitted work never leaves your machine. |
| Curl install | A tag-pinned one-liner; the tag in the URL is the version you get. |
More source-control providers¶
GitHub is the only supported remote today. These are on the way:
| Provider | Status |
|---|---|
| GitHub | Supported |
| Azure DevOps | Upcoming |
| GitLab | Upcoming |
| Forgejo | Upcoming |
| Others | Under consideration |
Until then, run against a GitHub remote or stay git-only with [scan].forge = "off". See
Providers.
Deferred capabilities¶
Larger, net-new pieces that aren't built yet:
- Cross-repo queries - the portal holds many projects, but each is still analyzed on its own; questions that span repos are not built.
- A multi-user, remote portal - local mode is single-user and loopback-only. A production mode with logins, sessions and organizations on their own subdomains, whose projects are imported from GitHub through a GitHub App, ships in 2.0.0. Agents reach its projects through a developer's local WhyGraph, but the production portal itself has no MCP endpoint: agents that talk to it directly (per-agent tokens, cloud agents) are what unlock the full service model beyond one machine.
- Per-branch CodeGraph index - the code index is still single-branch, so switching branches and re-syncing rewrites it. WhyGraph's own database no longer needs this: it keeps one database and computes branch membership per commit, recomputed and self-healing on every scan.
Want to weigh in?
These are shaped by what people actually need. Open an issue on GitHub if one of these matters to you.