Software intelligence for you and your coding agents.
AI is writing code faster than you and your team can understand it. SourceAnt helps you stay ahead. It shows you what is being built, checks changes against your decisions and requirements, and keeps your system architecture visible as the code moves.
Let agents move fast without losing control of where the software goes.
Run it on your computer now. No account required.
Out of the box, on your own machine:
- Code graph. Every repository you add, parsed into files, symbols and the relationships between them.
- Local code review. Read a checkout through the same reviewer a pull request goes through.
- Review on pull requests. The same reviewer on the forge, commenting where the line is.
- Knowledge and requirements. What you decided and what the software must do, kept between sessions and linked to the code that carries it.
- Skills over MCP. Choose and apply your skills from coding tools that support MCP prompts.
- One place for every index. All your repositories in a single local store, each under its own scope, and nothing written into your folders.
- Review across repositories. A change in one repository read against the others it reaches.
Install the CLI.
curl -fsSL https://raw.githubusercontent.com/sourceant/cli/main/scripts/install.sh | shsourceant setupReview the current checkout. Prepare your work for review by collaborators, improve quality, cut review time, check your work against rules, skills and workflows.
sourceant reviewThree commands to a first review. The rest:
sourceant repo add <path> |
Add a repository and read it into the graph |
sourceant repos |
What is added, and when each was last read |
sourceant graph <repository> |
What the indexer found in one repository |
sourceant architecture |
Components and what they depend on |
sourceant ui |
Open the graph and the reviews in a browser |
sourceant mcp |
Connect an MCP client over standard input and output |
sourceant status |
Whether the agent and the indexer are running |
sourceant stop |
Stop the agent and its core |
A client that speaks HTTP goes to http://127.0.0.1:8930/mcp instead, and reads
the same graph a review reads.
Docs: sourceant.ai/docs.
SourceAnt parses your repositories into one graph and serves it over MCP, so your coding tools read what its reviews read. Knowledge you record sits in that graph beside the code it governs, so a review of one file finds the decision that constrains it.
| Code | search_code, trace_code |
| Local indexing | get_index_status, index_repository |
| Skills | search_skills, get_skill, save_skill, delete_skill |
| Knowledge | put_knowledge, search_knowledge, put_knowledge_relationship |
| Requirements | put_requirement, link_requirement, search_requirements, get_requirement_coverage |
| System | put_topology_entity, put_topology_relationship, traverse_topology |
| All of it, in one pack | get_context |
You ask for it in plain language:
Remember that project shop uses signed webhook requests. Store it as an approved decision.
Connect the signed webhook decision to the rule that rejects unsigned requests.
Get the approved knowledge related to the signed webhook decision before changing its handler.
A change can break a repository it never mentions. Rename a route, and the client that calls it breaks under the old name. That name is nowhere in the diff.
SourceAnt records how your repositories are joined, and reviews across that join. A review searches each repository the change reaches for what would break there. A repository it could not read is named rather than skipped.
The joins come from two places. A manifest declares a dependency. A URL one service calls is declared nowhere, so a model reads the code for those and proposes them, quoting the lines. You approve a proposal before it counts.
The core is MIT licensed and self-hostable in full. SourceAnt Cloud runs the same engine with a managed layer on top: memory your team curates, contract analysis, continuous indexing at scale, the explorable graph, workspaces and roles, and analytics.
| License | Contact | Maintainer | Contributing |
|---|---|---|---|
| MIT, copyright Whilesmart LLC | [email protected] | WhileSmart | CONTRIBUTING |