This is the public roadmap for Codevira MCP itself — what's been built, what's coming next, and what's being considered for the future.
Have a suggestion? Open a feature request.
The foundation. Everything needed to give AI agents persistent memory on any project.
- 26 MCP tools across graph, roadmap, changeset, search, and code reader modules
- Context graph — YAML-based file nodes with rules, stability, blast-radius BFS
- Semantic code search — ChromaDB + sentence-transformers, fully local
- Roadmap system — phase-based tracker with auto-stub on first use
- Changeset tracking — atomic multi-file change management with session resume
- Decision log — structured session logs, searchable by future agents
- Agent personas — Orchestrator, Planner, Developer, Reviewer, Tester, Builder, Documenter
- Config-driven setup —
config.yamlfor watched dirs, language, collection name - Auto-reindex git hook — incremental re-index on every commit
- Works with Claude Code, Cursor, Windsurf, Google Antigravity, any MCP-compatible tool
Extend full feature support beyond Python.
- Tree-sitter integration — AST-based chunking for TypeScript, Go, and Rust
get_signaturefor TypeScript/Go/Rust — symbol extraction via tree-sitter- Auto-generated graph stubs for all languages — not just Python
- Language-specific playbook entries — task rules for TypeScript patterns, Go idioms
Make setup and daily use smoother.
- Progress bar for indexing — visual feedback during
--fullbuilds on large codebases codeviraCLI — single entry point:codevira init,codevira status,codevira index- Global Installation Support — capability to install codevira globally (e.g., via
pipxoruv) and use the MCP server across all projects without per-project setup - Index health dashboard —
codevira statusshows stale files, graph coverage, last indexed - VS Code extension — one-click MCP server setup without editing JSON config manually
Make the context graph richer and more automatic.
- Cross-file rule inference — auto-detect invariants from patterns across sessions
- Graph diff on PR — show what graph nodes changed in a pull request
- Dependency edge auto-refresh — re-derive
connects_toedges when imports change - Graph visualization — export to Mermaid or DOT format for documentation
Ideas being evaluated — not yet committed.
- JSONL interoperability for session logs — enable cross-framework agent integrations (mmkr, hydra, netherbrain etc.) by writing a parallel
.jsonllog alongside the existing YAML format - Multi-repo support — federated graph across multiple repositories in a workspace
- Remote ChromaDB — option to share a single index across a team (instead of per-developer local)
- GitHub Actions integration — run index + graph refresh in CI on every merge
- Natural language graph queries — "show me all files that publish events" via semantic graph search
- Plugin system — custom MCP tools per project without forking Codevira
Features move from "Considering" to a versioned milestone based on:
- Community demand — upvote issues or comment on feature requests
- Contribution — if someone opens a well-scoped PR, it gets prioritized
- Project fit — does it make AI-assisted coding better without adding fragility?
The roadmap is updated as milestones are completed or priorities shift.