# dc tools

> Deterministic context for AI-assisted development. dct (available now) is persistent memory and work tracking for AI coding agents — issues, todos, plans, sprints, decisions, changelogs and session handoffs in one local database the agent reads and writes over MCP. dcg (in development) parses a codebase into a queryable graph. Facts instead of guessing: same query, same answer.

Key facts:

- Install dct: `uv tool install 'dctracker[mcp]'` (add `,pg` inside the brackets for a PostgreSQL backend) then `dct init`
- dct v1.4.0 — Python 3.11+, SQLite by default (first-class PostgreSQL), MIT license
- The 52 MCP tools and the CLI work with any MCP client (e.g. Cursor, Windsurf); the 13 skills and 4 hooks are Claude Code integrations
- Any MCP client: register `dct server` (stdio) in the client's MCP config; in Claude Code `dct init` does it for you
- Local-first: your tracking data lives in a database on your machine and is never uploaded; the only default network call is a once-a-day PyPI version check (opt-out), and optional plan classification uses your own claude CLI
- Core workflow (plain conversation; the agent calls MCP tools): create_item (/track in Claude Code) → add_checkpoint / complete_checkpoint as acceptance criteria → add_changelog before commits (in Claude Code an on-by-default hook blocks agent commits without an entry) → handoff_work at session end → pickup_work to resume the next session from recorded state in a single read
- Plans: markdown plan/spec/roadmap/ADR files stay the source of truth; dct anchors their sections and checkpoints with stable ULIDs — plan state is queryable and promotable to items and sprints
- dcg — Deterministic Context Graph: codebase parsed into a graph database (Go single binary, Neo4j storage, Python and Go parsers, Django-aware), queryable over MCP; 17 tools; not yet released

## Docs

- [dct repository](https://github.com/fotodeveloper/dct): source, README, issue tracker
- [dct documentation](https://github.com/fotodeveloper/dct/tree/main/docs): guides and reference
- [dctracker on PyPI](https://pypi.org/project/dctracker/): package releases

## Optional

- [llms-full.txt](https://dctools.dev/llms-full.txt): full site content as markdown