About Ganty MCP
Ganty MCP connects AI clients to Ganty projects, tasks, and milestones through its hosted Gantt chart service. It also computes critical paths and schedule shifts server-side, returning dependency-aware results rather than relying on the LLM for date calculations.
Key features
- 01List workspaces, projects, and tasks
- 02Create and update project data
- 03Compute critical paths and slack
- 04Simulate or commit task rescheduling
- 05Propagate shifts across dependencies
Use cases
- 01Inspect tasks and project milestones
- 02Find a project's critical path
- 03Estimate release impact of a task delay
- 04Update schedules while preserving pinned tasks
MCP support
The hosted Streamable HTTP MCP server exposes read and write tools for Ganty. Bearer tokens have read or write scope, and calculations such as critical-path analysis run server-side.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://ganty.app/api/mcpclaude mcp add --transport http app-ganty https://ganty.app/api/mcp{"mcpServers":{"app-ganty":{"type":"http","url":"https://ganty.app/api/mcp"}}}Open in Cursor{"servers":{"app-ganty":{"type":"http","url":"https://ganty.app/api/mcp"}}}Open in VS Code{"mcpServers":{"app-ganty":{"type":"http","url":"https://ganty.app/api/mcp"}}}{"mcpServers":{"app-ganty":{"serverUrl":"https://ganty.app/api/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
15This tool takes no parameters.
Details
From the project READMEGanty MCP
Ganty is a Gantt chart SaaS with a built-in Model Context Protocol (MCP) server that lets AI clients like Claude Desktop and Claude Code drive your projects in natural language — and computes critical paths and shift-impact server-side rather than asking the LLM to do math.
Status: v1.0.0 — production endpoint at
https://ganty.app/api/mcp(Streamable HTTP) Free on every Ganty plan including Free.
What makes this different
Most MCP servers expose CRUD over an existing API. Ganty's also exposes calculation tools that compute on the server and return JSON — so the LLM never has to reason about dates, dependencies, or critical paths.
get_critical_path— forward/backward-pass CPM with progress-aware remaining duration. Returns the critical path tasks in order, per-task earliest/latest start/finish, slack, projected end date. Detects cycles before computing and returnserror: 'cyclic_dependency'instead of hanging or guessing.reschedule_and_propagate— simulate or commit a task shift. Default mode isdry_run(no DB writes). Push-only cascade respects pinned tasks (anything atprogress=100or inpinned_task_ids) and reportsconflictsinstead of overwriting.commitmode wraps everything in an all-or-nothing transaction.
Quick start
- Issue an MCP token in Ganty: dashboard → Settings → MCP Integration → "Issue new token".
- Configure Claude Desktop (
claude_desktop_config.json) or Claude Code (.mcp.jsonat project root):{ "mcpServers": { "ganty": { "url": "https://ganty.app/api/mcp", "headers": { "Authorization": "Bearer YOUR_TOKEN" } } } } - Restart Claude and try a prompt:
"List my Ganty projects""What's the critical path for the e-commerce rebuild project?""If I push the payments task by +3 days, when does release ship?"
Full setup walk-through: https://ganty.app/guide/mcp-integration
Tools
Read tools (visible to read-only and write tokens)
| Tool | Description |
|---|---|
list_workspaces | Workspaces the token's user belongs to |
list_projects | Projects in a workspace |
list_tasks | Tasks in a project (filter by name/status/assignee) |
get_task | Single task with assignees + dependencies |
list_milestones | Milestones in a project |
get_critical_path | Server-side CPM with slack and total duration |
Write tools (write-scope tokens only)
| Tool | Description |
|---|---|
create_project, delete_project | Project lifecycle |
create_task, update_task, delete_task | Task lifecycle |
set_task_progress | Update task progress (auto-sets status) |
add_dependency | Finish-to-Start dependency between two tasks |
create_milestone | Add a milestone |
reschedule_and_propagate | Shift a task, cascade through dependencies, dry-run or commit |
reschedule_and_propagate is classified as a write tool even in dry_run mode (it shares the same compute path as commit and we don't want read-only tokens running unlimited simulations).
v1 scope and limitations
We prefer explicit non-support over silent wrong answers. Every response from the calculation tools includes a limitations array.
- Dependency types: Finish-to-Start only (Ganty's schema doesn't model SS/FF/SF)
- Calendar: calendar days by default;
business_days: trueskips Sat/Sun only (no holiday table) - No resource calendars in v1
- Multi-period tasks (extraSegments) are ignored in v1; calculations use only the main start/end of each task
- Pinned tasks:
progress=100automatically pinned; supplypinned_task_idsfor additional pins - Push-only cascade: successors are pushed back when required, never pulled forward
Architecture
- Transport: Streamable HTTP (JSON-RPC 2.0 over HTTP POST) —
https://ganty.app/api/mcp - Auth: Bearer tokens (
gnty_...) issued from the Ganty dashboard. Tokens are scopedreadorwrite. - Rate limits: 120 req/min for read, 30 req/min for write (per token)
- Audit logging: every write call is logged with action, actor, IP, and arg summary
- Atomicity:
reschedule_and_propagatecommitmode is wrapped in a Prisma transaction; conflicts abort without writes
The calculation logic lives in pure TypeScript functions backed by 28 golden tests asserting concrete dates and day counts (linear chain shifts, diamond merges, cycle detection within 2 seconds, progress-aware remaining duration, hand-calculated CPM with slack, business-day mode, lag-day handling, push-only cascade).
Links
- Marketing page: https://ganty.app/integrations/mcp
- Setup guide: https://ganty.app/guide/mcp-integration
- Claude × Gantt guide: https://ganty.app/blog/claude-gantt-chart
- Design article: https://ganty.app/blog/mcp-server-side-calculation-engine
- Tools by example: https://ganty.app/blog/mcp-gantt-chart-5-examples
- Troubleshooting: https://ganty.app/blog/mcp-setup-troubleshooting
- Sign up (free 5-member plan): https://ganty.app/signup
License
MIT for this documentation repository. The Ganty server itself is a hosted SaaS; the MCP endpoint is a feature of every Ganty workspace.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.