Redmineflux MCP is live - connect any AI agent to Redmine.Explore More →
For Redmine teams who want AI without a per-seat subscription bill.
Why this matters
One Anthropic or OpenAI API key configured once in Redmineflux – every teammate uses the in-product chatbot through it. No per-seat subscriptions. No licence sprawl. No 500-engineer Slack thread about who has Claude Pro this month.
How it works
Your team’s questions become MCP tool calls. The Redmineflux MCP server resolves them against the same Redmine your team has always worked in. Plugin-aware, audit-logged, plain-English.
The MCP server runs as a managed service, bundled with every Redmineflux Cloud Pro subscription — zero install, zero maintenance.
In plain words
The MCP server is the bridge between Redmineflux and any AI agent. Two ways to use it: power users connect their existing tools – Claude Desktop, Claude Code, ChatGPT – directly to the MCP server. Everyone else uses the Redmineflux chatbot, built into every page of the product. Same MCP server underneath. One Anthropic or OpenAI key for the whole team – no per-seat subscriptions, no licence sprawl.
The Model Context Protocol is an open standard from Anthropic. It lets AI agents connect to external systems through a structured, secure interface – without custom integrations for every AI tool. Redmineflux speaks it natively.
Redmine holds the data. Redmineflux enriches it. The MCP server makes it readable, writable, and auditable to any AI agent your team already uses. Same Redmine your team has always used – now able to answer “how many bugs are open in Phoenix?” in seconds.
What it actually feels like
Production-tested against the NovaCrest dataset – 3,500 issues, 22 users, 6 projects. Below is one real session, end to end.
What should I work on first this morning?
workload.next_priority(user="me", limit=5)
You have 264 open issues across 4 projects. Three at Immediate priority assigned to you:
Take #2972. Log 4 hours of dev, mark it 80% done. Comment: “API refactoring complete.”
issue.update(id=2972, percent=80) && time.log(issue=2972, h=4, note="API refactoring complete")
Done. Logged 4.0 h on #2972, status now 80% complete. Audit row written.
What “talking to your Redmine” actually looks like
Four common asks. The MCP call your AI agent makes. The outcome the human sees.
“Who’s overloaded this week?”
workload.over_capacity(team="eng", week="current")
Three engineers flagged at 38–42h booked against 35h capacity. Top issues per person, ready to rebalance before standup.
“Log 4 hours on issue #50, mark it 80% done.”
issue.update(id=50, percent=80) && time.log(issue=50, h=4)
Logged in 3 seconds. Permission-checked against the authenticated user. Audit row written before the response returns.
“What ships in the Phoenix sprint, and what’s at risk?”
project.summary(slug="phoenix", include=["risk","milestones"])
Sprint summary in plain English. 12 issues on track, 3 ageing, 1 blocked on a vendor decision. PM gets the status update without filtering anything.
“Show me every AI action in the last hour.”
tail -f logs/mcp-session-*.jsonl | jq 'select(.kind=="write")'
Every tool call as a JSON Lines record – session, user, action, parameters, timing, result. Pipe to your SIEM. Stream to your observability stack.
No noise. No fake capabilities. The server auto-detects which Redmineflux plugins are installed on your Redmine instance and only exposes what your installation actually supports. Nothing your AI can hallucinate against.
Compared to open-source Redmine MCP servers
| Capability | Open-source Redmine MCP servers | Redmineflux MCP server |
|---|---|---|
| Issues (list, create, update) | Yes | Yes |
| Time tracking | Basic | Full – log, list, filter by date / project / user |
| Cross-project queries | No | Yes – one query, all projects |
| Convenience tools (workload, stats, onboard) | No | Yes – designed for real human queries |
| Plugin-aware capabilities (Timesheet, Workload, Agile, Knowledge Base, more) | No | Yes – capability injection per installed plugin |
| Audit logs (every tool call → JSON Lines) | No | Yes – session, user, params, timing |
| Capability injection (auto-detect installed plugins) | No | Yes |
| Hosted endpoint (zero install) | No | Yes — included in Cloud Pro |
Atlassian Rovo locks you to Atlassian’s managed model. Easy8 locks you to Easy8’s. Redmineflux locks you to nothing. Pick your provider – Anthropic, OpenAI, Azure OpenAI, a local Ollama runtime. Your data goes where you say it goes. EU and JP residency are first-class, not an upgrade tier.
Deployment
Audit + observability
Every tool call is logged as JSON Lines: session ID, authenticated user, tool name, parameters (sensitive fields redacted), timing, and result status. Stream them to your SIEM, your compliance store, or your observability stack.
For regulated buyers — banking, manufacturing, aerospace, cybersecurity — that means: every AI interaction has a paper trail, API key authentication is enforced, EU and JP data residency available, and no prompt content is stored unless you opt in.
FAQ
Yes. Production-tested against a live 3,500-issue Redmine dataset. Included in every Redmineflux Cloud Pro subscription — no separate licence, no install required.
The MCP server makes Redmine accessible to external AI agents – Claude, ChatGPT, Copilot, anything you already use. The Redmine AI Assistant (inside the Redmine UI) is a separate product, rolling out with our next release. Both use the same MCP server underneath.
Any agent that supports the Model Context Protocol. Verified today: Claude Code and Claude Desktop (Anthropic), ChatGPT with MCP enabled (OpenAI), GitHub Copilot (Microsoft), and any custom agent built on the MCP TypeScript or Python SDK.
The MCP server is currently available exclusively as part of Redmineflux Cloud Pro. If you are on self-hosted Redmine and want MCP access, contact us to discuss options.
The MCP server auto-detects which Redmineflux plugins are installed on your Redmine instance. It loads only the tools those plugins actually support. If you don’t have Workload installed, the server doesn’t expose Workload tools. No fake capabilities, no noise.
Every tool call writes a JSON Lines record to the logs/ directory: session ID, user, tool name, parameters (sensitive fields redacted), timing, result. Ship them to your SIEM or compliance system. No prompt content is stored unless you opt in.
Plugin-aware capabilities for Agile Board, Knowledge Base, and Custom Dashboard rolling out with our next release. CRM, Helpdesk, and Test Case Management coming this season. An MCP Dashboard plugin that visualises agent activity inside Redmine – in trial with early customers.
