tkmd
A local Markdown task manager and program engine that uses CPM scheduling and verification gates to govern developers and AI coding agents.
At a Glance
Self-hosted local CLI, dual-licensed under MIT or Apache-2.0 at the user's option. No cloud accounts or subscriptions.
Engagement
Available On
Alternatives
Listed Oct 2026
About tkmd
tkmd (CLI: tk / tkmd) is a Rust command-line task and project manager built by Ken Harrison (ktreharrison). It stores each task as a plain Markdown card with YAML frontmatter, schedules work with Critical Path Method and Design Structure Matrix algorithms, and acts as a governance harness for AI coding agents. The latest listed release is v0.1.2.
What It Is
tkmd is a deterministic program engine for developers and AI agents. Tasks live as tasks/tk-XXX-*.md files, and state transitions are recorded in an append-only .events.jsonl stream. It runs locally as a single binary, and the README states it makes no network requests and has no telemetry.
How the Agent Governance Works
Agents claim tasks (tk c), attaching a session ID so the conversation can be resumed later (tk u). Before submitting work, an agent must pass the card's shell eval command, then move the task to pending-close (tk r). Only human sign-off (tk d) moves a task to done, which then auto-promotes downstream blocked tasks to ready. POSIX advisory file locking prevents write collisions between concurrent agents.
Scheduling and Analysis
tk dep --cpm computes early start, late finish, schedule buffers and the critical path. tk dep --dsm detects circular dependencies, and tree, Gantt and AON views are also available. tk st computes flow telemetry such as Little's Law, MTTC and TPM90 natively, and the JSONL event log can optionally be queried with DuckDB.
Setup and Storage Modes
Install via a shell script or cargo install tkmd. tk init supports an Obsidian mode that syncs with a .base Kanban board and Daily Notes, a standalone CLI mode, and a per-repository local backlog (tk init -L). Local, global and merged scopes can be selected with flags. Shell completions are generated for Zsh, Bash, Fish, PowerShell and Elvish.
MCP and Agent Skill
tkmd mcp runs a JSON-RPC 2.0 MCP server over stdio. The README lists compatibility with Claude Desktop, Codex, Gemini CLI, Cursor, Zed, OpenCode and Pi. A bundled agent skill in skills/tkmd enforces the review-gate workflow.
Community Discussions
Be the first to start a conversation about tkmd
Share your experience with tkmd, ask questions, or help others learn from your insights.
Pricing
Open Source
Self-hosted local CLI, dual-licensed under MIT or Apache-2.0 at the user's option. No cloud accounts or subscriptions.
- Dual-licensed: MIT or Apache License 2.0 (user's choice)
- 100% local and offline, zero network requests, no telemetry
- Markdown task cards with YAML frontmatter
- CPM and DSM scheduling (tk dep --cpm / --dsm)
- Agent verification gates (eval) and pending-close human sign-off
Capabilities
Key Features
- Markdown task cards with YAML frontmatter
- Critical Path Method scheduling (tk dep --cpm)
- Design Structure Matrix cycle detection
- Eval shell verification gates
- pending-close human sign-off workflow
- Automatic blocker cascade unblocking
- Session resume for AI agent sessions
- Stdio MCP server
- Obsidian Base Kanban sync
- Local, global and merged scopes
- Append-only JSONL event log and flow telemetry
- Fuzzy task search
- Shell completions
- Bundled agent skill
