kanban-mcp
kanban-mcp is an MCP (JSON-RPC 2.0 over STDIO) server that provides a persistent, database-backed kanban board (issues/features/todos/epics/diary entries) with agent-facing tools (40+), optional semantic search via local ONNX embeddings, an activity timeline, export (JSON/YAML/Markdown), and a web UI (localhost:5000).
Score Breakdown
⚙ Agent Friendliness
🔒 Security
TLS assessment is for the web UI only; MCP is STDIO/local process, and no network auth/TLS enforcement details are stated. Database credentials are stored in a local .env file (with precedence rules). Dependency hygiene and secret-in-logs assurances are not verifiable from the provided README. No user-level authz/role model described; treat it as trusted-local software or add external controls if exposing beyond localhost.
⚡ Reliability
Best When
You want local/offline or self-hosted agent coordination with persistent state and can run an MCP client over STDIO (e.g., Claude/Gemini/Copilot/VS Code MCP integrations).
Avoid When
You need robust authz (per-user permissions), API rate-limit guarantees for untrusted callers, or you cannot accept local storage and optional model download on first semantic-search use.
Use Cases
- • AI coding agents tracking project work items across sessions
- • Keeping a persistent backlog→done workflow with status history and relationships (epics/parent-child/blockers)
- • Semantic search over local kanban items using ONNX embeddings
- • Generating reports/exports for planning and reviews
- • Human/agent collaboration via a local web UI while agents interact through MCP
Not For
- • Use as a multi-tenant hosted SaaS for sensitive data (appears primarily local/hosted by the user)
- • Public internet exposure without additional access controls
- • Workflows requiring OAuth/SSO-based enterprise auth or fine-grained user identity/permissions
- • High-reliability enterprise service guarantees without operational/monitoring setup
Interface
Authentication
Authentication described is for database access (MySQL user/password or local SQLite). The MCP interface itself appears to rely on local process trust and client-provided env, without user-level auth/authorization details in the README.
Pricing
No pricing model described (open-source/local deployment implied).
Agent Metadata
Known Gotchas
- ⚠ Semantic search first query may block while downloading the ONNX embeddings (~140MB)
- ⚠ MySQL setup may involve interactive prompts and root password requirements for non-interactive/agent usage
- ⚠ MCP runs via STDIO and spawns as a subprocess; ensure the MCP client environment inherits or supplies KANBAN_DB_* variables if needed
- ⚠ State is persisted in local DB; agents should avoid unintended duplicate creates without checking existing items (no idempotency guidance provided)
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for kanban-mcp.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-30.