QodFlow — MCP server for Claude Code: install and get started
MCP server that connects AI agents to QodFlow, a kanban board with SLA tracking shared by people and AI. Agents claim jobs, report progress, move them between stages, and ask for human approval when needed.
MIT updated 2026-10-29
claude mcp add qodflow-mcp -e QODFLOW_API_TOKEN=<QODFLOW_API_TOKEN> -- npx -y qodflow-mcpWhat QodFlow does
QodFlow exposes 27 tools over MCP: create and list jobs, move them across pipeline stages, manage assignments and checklists, add timeline notes, request human decisions with voting or options, and acknowledge approvals. The agent retrieves a briefing on startup showing pending decisions, unread approvals, and expired requests that block progress.
Setup requires a Personal Access Token generated in QodFlow (Settings → API Keys → New token). Pick the scopes you need; read-only work starts with jobs:read and stages:read. Pass the token via the QODFLOW_API_TOKEN environment variable. The server connects to https://www.qodflow.com via MCP, keeps nothing on disk, and enforces a 120 requests/minute rate limit per token. Destructive operations (team deletion, mass moves) are never exposed through MCP; they live behind the UI's confirm-literal guard only.
One nuance: when the agent requests human approval and you give it in chat, use the tool to report that approval back to the board. Board-level approvals are recorded separately and show as verified; chat approvals show as reported by the agent.
What it helps with
- Let AI agents claim and complete jobs on a shared kanban board without UI switching
- Request human approval for a decision in chat and have the agent record it on the job
- Track job age at each stage and enforce SLA compliance across the team
- Spawn jobs from templates and fill checklists as work progresses
- Give agents visibility into the job queue, blockers, and pending decisions
Example prompts for the agent
- Create a job from the standard template and move it to done
- Show me all active jobs tagged urgent and ask which ones I want to approve for release
- Request a decision: should we move this job to completed?
How to install QodFlow
Platform CLI
claude mcp add qodflow-mcp -e QODFLOW_API_TOKEN=<QODFLOW_API_TOKEN> -- npx -y qodflow-mcpReplace <…> with your own values
Platform CLI
codex mcp add qodflow --env QODFLOW_API_TOKEN=<QODFLOW_API_TOKEN> -- npx -y qodflow-mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"qodflow": {
"command": "npx",
"args": [
"-y",
"qodflow-mcp"
],
"env": {
"QODFLOW_API_TOKEN": "<QODFLOW_API_TOKEN>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"qodflow": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"qodflow-mcp"
],
"env": {
"QODFLOW_API_TOKEN": "<QODFLOW_API_TOKEN>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e QODFLOW_API_TOKEN=<QODFLOW_API_TOKEN> qodflow npx -y qodflow-mcpReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about QodFlow
How do I install QodFlow for Claude Code?
Run claude mcp add qodflow-mcp -e QODFLOW_API_TOKEN=<QODFLOW_API_TOKEN> -- npx -y qodflow-mcp. Replace <QODFLOW_API_TOKEN> with a token from QodFlow (Settings → API Keys → New token).
What can the agent do through QodFlow?
Create, move, and assign jobs; claim and release jobs; add checklists and progress notes; ask humans for approval; read the board with filters by stage, assignee, tag, and due date; track how long jobs spend on each stage.
What token do I need and where do I get it?
Sign in to QodFlow, go to Settings → API Keys, click New token, select the minimum scopes (for read-only use jobs:read and stages:read; add jobs:write to create and edit), copy the token once (it will not appear again), and pass it to the MCP server as QODFLOW_API_TOKEN in your config.