Describe a project and the Wizard turns it into a task board with its own knowledge base. Assign any task to MFlow's coding agent — your own OpenAI, Claude, Gemini, or DeepSeek key — and it ships code inside an isolated sandbox, reading your project's docs and conventions first.
Free tier · No credit card · Connects in under 2 minutes
Also — commented a summary back on WR-42, so anyone watching the board sees what shipped.
Sandbox
Context read first
Trusted by product & engineering teams
Most "AI writes code" demos are a landing page finished in thirty seconds. MFlow's coding agent is built for the project you're already maintaining — the one with a hundred files, a dependency tree, and conventions only your team knows.
The Wizard turns a spec into a task board and a project knowledge base. That board and KB are the shared context every agent — yours or ours — reads from next.
From the board itself, or from any MCP client — Claude Code, Claude Desktop, claude.ai — reference a task and hand it off. No copy-pasting a prompt between tools.
Inside its own isolated, self-hosted sandbox, the agent reads the task, the project's knowledge base, and your repo's own convention files — AGENTS.md, CLAUDE.md, GEMINI.md — before touching a line of code.
Clones your GitHub repo, works in a sandbox sized for the job, and commits back — sized for a real codebase with real dependencies, not a five-file demo.
What it actually does
Use MFlow's hosted key, or connect your own — sandbox starts are metered by your plan either way.
Connect Claude Code, Claude Desktop, or claude.ai directly to your MFlow account via the Model Context Protocol. Ask about your projects, create tasks, or check delivery risk without leaving your terminal or editor — no API key, no client secret, just sign in through the same MFlow account you already have.
See the full MCP setupOne click spins up a personal, isolated sandbox — a full copy of MFlow, sample projects, and the AI assistant included. No platform connection, no account, and nothing about you is stored. Explore it exactly as your team would, then decide.
Your sandbox provisions automatically in a few seconds. Fair-use rate limits apply to keep it fast for everyone.
MFlow stores your Jira, Trello, and Standalone tasks, project data, and activity history — that's what powers the metrics and AI. We don't touch comments, attachments, or anything outside delivery data.
Your data is processed lawfully under GDPR. We never sell it, share it with third parties, or use it to train models. It exists solely to run your analytics.
From your account panel, one click permanently erases all your data from MFlow — Jira, Trello, and Standalone alike. No support ticket, no waiting. Most SaaS tools make you ask. We just let you do it.
No setup, no custom fields. Connect Jira or Trello — or start fresh in Standalone — and get your sprint status, risks, and AI assistant ready in minutes.
Start for free →One price per user. Sandbox starts and concurrent coding tasks are enforced by your plan.
Free
free forever
For trying MFlow and the coding agent on real work — one Jira project, light and mid sandbox tiers.
Pro
per user, billed monthly
For builders who need more projects, more sandbox starts, and two coding tasks running at once.
Scale
per user, billed monthly
For fast-growing teams that need the highest sandbox quotas and three concurrent coding tasks.
Custom
contact for pricing
For security-first organizations that need custom deployment, data, or support terms.
Talk to salesSandbox starts are counted per calendar month and apply whether the coding agent uses MFlow's hosted model or your own LLM key. When the sandbox host is busy, or you reach your plan's concurrent-task limit, new coding tasks wait in a queue instead of failing — a queued task can be cancelled and does not use a sandbox start until it actually begins.
No credit card required for the free tier · Cancel anytime