Sigmap

SigMap is a command line tool that builds a deterministic, auditable map of a codebase for AI coding assistants. It extracts code signatures and evidence from a project and uses them to decide which files an assistant should see for a given question, with no embeddings and no native dependencies, running fully offline. The workflow centers on a few commands. Running npx sigmap generates context for the project, ask returns a ranked list of relevant files and a ready to paste context file, validate checks coverage, judge scores how well an answer is grounded in the supplied files, and learn reinforces the files that helped. It supports 38 languages including TypeScript, Python, Go, Rust, Java, Kotlin, Swift, C#, SQL, Terraform and R. It also provides an MCP server for on demand context lookups, and works with Copilot, Claude Code, Cursor, Windsurf, Codex, OpenCode and Gemini CLI, as well as local models run through Ollama, llama.cpp and vLLM. The vendor publishes benchmark results: 80.4 percent hit at five across 18 repositories and 105 tasks, a 44.4 percent reduction in prompts needed to finish a task, and a 95.7 percent average token reduction, along with a self contained HTML report. It is aimed at developers and teams who want smaller, more trustworthy context for coding agents, including in CI and monorepos. No pricing is published. The tool is run directly through npx and the site is a documentation site with a getting started guide.