Connect
An approved MCP client connects using an identity and tool scope defined by administrators.
Bunny Redmine by Appar
Redmine MCP Server
Connect Cursor, Claude Code, Codex, and other AI agents to authorized Redmine projects, tools, and knowledge through a governed MCP server.
Model Context Protocol for Redmine
Give AI agents real Redmine context and controlled write access
Bunny Redmine MCP replaces temporary Markdown handoffs with live, permission-aware project context. Agents can read, reason, act, and write results back to the system your team already trusts.
Execute
The agent reads authorized context, calls only available tools, and carries out the assigned work.
Record
Every tool call, modification, result, cost, and error is logged and written back to Redmine.
Redmine MCP tools
More than issues: expose the Redmine workspace AI agents need
Bunny can provide a carefully scoped tool surface instead of giving every agent a broad administrator token.
Illustrative product walkthroughA coding agent calls only the Redmine tools it can use.
Illustrative product walkthrough
An approved MCP client connects with its assigned identity.The server exposes only the account's permitted data and selected tools.The external agent calls the available tools for the task.The execution result and source context remain in Redmine.
Issues and relations
Search, create, update, comment, assign, relate, close, and organize parent or child issues.
Projects and versions
Read or create projects, subprojects, memberships, versions, milestones, and categories.
Wiki, News, and Forum
Read and publish structured knowledge, decisions, reports, updates, and discussions.
Files and attachments
Upload deliverables, inspect supported files, and attach screenshots, recordings, PDFs, or test evidence.
Time and workload
Query time entries, workload, ownership, due dates, and delivery signals before acting.
Custom enterprise tools
Add approved internal systems without exposing unnecessary tools to every employee or agent.
Cursor and Claude workflow
From a Redmine issue to parallel AI execution and verified delivery
Claim structured work
The coding agent receives an authorized Redmine task with requirements, files, history, and acceptance criteria.
Build and test
Cursor, Claude, Codex, and browser agents work in parallel and attach code, screenshots, recordings, and test results.
Update and approve
The agent updates the issue and related documentation; a human reviews the evidence and makes the final decision.
Secure enterprise MCP
Writable MCP without uncontrolled access
Permissions, exposed tools, approval points, logs, models, and cost are managed explicitly for each account and AI Employee.
ADMINISTRATOR CONTROLSGive AI a clear scope. Keep every action traceable.
AI EmployeeProject assistant
AI can access only what its Redmine account can access.
Only the tools you allow are available to this Employee.
At the monthly limit, the next job stops.
See what ran, what was blocked, and why.
Bound Redmine accountProject manager · pm-teamSame project permissions as this account
Permitted projectAccessible
Restricted projectNot accessible
Tool allowlist
Only selected tools are available to this AI Employee.
create_issueAccessiblemanage_wikiAccessiblecreate_projectNot accessibleHard limit$50 USD
Budget reached. Execution stopped.
Start next scheduled jobBudget stop
Execution logs.json
ActionResult
Read permitted issueCompleted
Read restricted projectPermission denied
Start next scheduled jobBudget stop
View example JSON record
{
"example": true,
"account": "pm-team",
"tool": "get_issue_content",
"status": "completed",
"monthly_budget_usd": 50,
"next_execution": "budget_exceeded"
}Configured by system administrators
Illustrative product walkthrough
Bunny inherits the bound account's access. Restricted projects stay restricted.Administrators choose the tools each AI Employee can use.A monthly hard limit stops execution when the configured budget is reached.Review the outcome and execution context in a structured JSON record.
Account-bound permissions
An agent inherits the project visibility and permissions of its assigned Redmine account.
Minimal tool surface
Administrators expose only the tools needed for the employee, workflow, or MCP client.
Write controls
Read-only, direct-write, and human-confirmation behavior can be defined per action.
Structured audit log
Tool inputs, outputs, changes, errors, tokens, cost, and execution time are retained.
Customer-provided keys
Keep model credentials under customer control and use cloud or local models.
Kill switch and budgets
Stop execution immediately or automatically when a monthly hard limit is reached.
FAQ
Redmine MCP Server FAQ
Which AI clients can connect?
Bunny is designed for MCP-compatible clients such as Cursor, Claude Code, Codex, and other approved agent platforms.
Can MCP write to Redmine?
Yes. Administrators decide which write tools exist and where human confirmation is required.
Does every employee receive the same tools?
No. Toolsets and permissions can be configured independently for each AI Employee.
Can it access Wiki, News, and Forum?
Yes. Bunny can control more than issues, including Redmine knowledge and communication areas.
Connect Redmine to AI agents
Give your next Redmine task to Cursor or Claude
Appar will help define the accounts, tool surface, approval rules, deployment, and agent workflow.