# SourceAnt

> SourceAnt connects your code, the services it depends on, and the decisions behind it. Bring that context into your agents’ work and your code reviews, so the next change builds on what your team already knows.

Code review, issue triage, Lens, and analytics are applications built on that intelligence. The engine is open source and MIT licensed. It indexes code locally, stores durable knowledge and system architecture, and speaks MCP. Requirements in the core are specialized knowledge records with links and coverage. SourceAnt Cloud adds managed memory, contract analysis, continuous indexing at scale, the explorable graph, workspaces and roles, hosted serving, and analytics. Premium requirements engineering will add a requirements graph, synchronization, change impact, and judged satisfaction. Plans band on repositories and analysed events, never per seat.

## Docs

- [Documentation](https://sourceant.ai/documentation): Entry point covering knowledge management on the cloud or the open core, code review, issue triage, self-hosting, and the plugin system.
- [Getting started](https://sourceant.ai/docs/getting-started): Sign up, connect your repositories and open a pull request, then reach SourceAnt from your editor.
- [Knowledge Management](https://sourceant.ai/docs/knowledge-management): How knowledge is stored as typed objects, how teams govern it, and the three ways to run it (cloud, open core, on-prem).
- [Cloud Setup](https://sourceant.ai/docs/cloud): Connect a repository and start capturing engineering knowledge on the hosted service, with no infrastructure to run yourself.
- [Local Setup](https://sourceant.ai/docs/local-setup): Explore code and knowledge, manage skills and check local changes against them.
- [Self-Hosting Guide](https://sourceant.ai/docs/self-hosting): Deploy the review API on your own server using Docker images.
- [Code Review](https://sourceant.ai/docs/code-review): Reviews on every pull request, judged against captured decisions, contracts, and blast radius rather than against the diff alone.
- [Settings](https://sourceant.ai/docs/settings): Configure models and keys, review behaviour, caches and MCP access.
- [Systems](https://sourceant.ai/docs/systems): Group repositories and resources, discover connections and review their evidence.
- [Systems review](https://sourceant.ai/docs/cross-repository-reviews): Group related repositories into a system so a review follows a change into the repositories it reaches.
- [Issue Triage](https://sourceant.ai/docs/repo-management): Automatic duplicate detection and labeling for issues and pull requests, grounded in the same knowledge graph that powers review.
- [Building a Plugin](https://sourceant.ai/docs/plugins): Extend the core with your own capture, enrichment, and automation, or swap in your own storage behind the knowledge and topology interfaces.
- [Contributing](https://sourceant.ai/docs/developer): Architecture, commands, and queue modes for contributing, and the storage-neutral core interfaces SourceAnt is built on.

## Product

- [Knowledge graph](https://sourceant.ai/knowledge-graph): Keep engineering decisions and constraints connected to the code and available to your tools.
- [Code review](https://sourceant.ai/code-review): Review changes with the decisions and context behind the diff.
- [Issue triage](https://sourceant.ai/issue-triage): Find potential duplicates and apply existing repository labels to incoming work.
- [Analytics](https://sourceant.ai/analytics): Track context coverage, review quality and token usage.

- [Code graphs](https://sourceant.ai/code-graphs): Build a central local graph of files, symbols and relationships, or import SCIP when another indexer covers the language better.
- [System Architecture](https://sourceant.ai/systems): Living architecture documentation that helps you steer change and gives your agents the connections they need before editing code.
- [Requirements Engineering](https://sourceant.ai/requirements): The open core records requirements, links, and coverage. The premium roadmap adds a requirements graph, synchronization, change impact, judged satisfaction, and a team workspace.
- [Use cases](https://sourceant.ai/use-cases): What teams do with a system record that stays current, each listed with the capability that actually ships behind it.
- [Self-maintaining architecture record](https://sourceant.ai/use-cases/architecture-record): An architecture record that tells you when it has aged, instead of a diagram that is quietly wrong.
- [Onboarding](https://sourceant.ai/use-cases/onboarding): A new engineer and a new agent session ask the same questions. Both are answered from the same records, so a senior stops being the interface to the codebase.
- [Contract drift](https://sourceant.ai/use-cases/contract-drift): A published interface versioned and compared against its baseline, joined to the services that declare a dependency on it.
- [CVE monitoring](https://sourceant.ai/use-cases/cve-monitoring): Advisories from the GitHub Advisory Database, checked against the code graph so a package nothing calls is not filed next to one that runs on every request.
- [Supply chain naming](https://sourceant.ai/use-cases/dependency-confusion): Which internal dependencies rest on a package name that carries no owner, read out of the confidence already on each edge.
- [How SourceAnt compares](https://sourceant.ai/comparison): Sourcegraph indexes code, graphify maps its structure, SourceAnt records what the team decided and whether it is still true.
- [Pricing](https://sourceant.ai/pricing): Bands on repositories and analysed events. Bring your own model keys, or use cloud models at cost plus a small margin.
- [SourceAnt Lens](https://sourceant.ai/lens): See the risk, evidence, and unresolved claims in a change before you approve it.

## Writing

- [Blog](https://sourceant.ai/blog): Notes on AI-assisted engineering, context, knowledge graphs, and code review.
- [Why AI suggests deleting code that matters](https://sourceant.ai/blog/why-your-ai-suggests-deleting-code-that-matters): Context-blind AI calls an intentional workaround dead code and removes the fix. Nothing remembers why the line is there.
- [How a knowledge graph cuts review tokens](https://sourceant.ai/blog/how-a-knowledge-graph-cuts-review-tokens): Reading whole files to answer a structural question is expensive. Traversing a typed graph and returning only the relevant subgraph is cheap.

## Links

- [Full docs as one file](https://sourceant.ai/llms-full.txt): Every page above, concatenated, for when the whole corpus fits in context.
- [Open core on GitHub](https://github.com/sourceant/sourceant): The MIT-licensed engine.
- [Website](https://sourceant.ai): Product overview.
- [App](https://app.sourceant.ai): SourceAnt Cloud.

## Optional

- [Terms of Service](https://sourceant.ai/terms)
- [Privacy Policy](https://sourceant.ai/privacy)