The MCP Connector lets AI assistants — Claude Desktop, Claude Code, Cursor, ChatGPT, or any client that speaks the Model Context Protocol — read and write your Navis Ops workspace directly. Ask your assistant to plan your week, triage your inbox, draft meeting follow-ups, or scaffold a new project, and it acts against your real data in real time.Documentation Index
Fetch the complete documentation index at: https://docs.navisops.com/llms.txt
Use this file to discover all available pages before exploring further.
Connect your AI client
You have two ways to authenticate: an API key (works with every MCP client, takes under a minute) or OAuth (cleaner for clients that support it, such as Claude Desktop and Cursor).- API key
- OAuth
API key authentication works with every MCP client. You create a token in Navis Ops and paste it into your client’s configuration file.
Create an API key
Click Create API key. Give it a recognizable name (for example, “Claude Desktop”) and choose a scope:
- Read — the assistant can query your data but cannot create, update, or delete anything.
- Read + Write — the assistant can take actions in your workspace.
Copy your token immediately
After creation, Navis Ops shows the full token exactly once. It starts with
nops_…. Copy it now — you cannot retrieve it later. If you lose it, revoke the key and create a new one.Configure your AI client
In your AI client’s MCP configuration, set the server URL and authorization header:The exact format of the configuration file varies by client. Check your client’s documentation for where to add MCP servers.
What the assistant can do
Your connected assistant has access to 52 tools across 12 areas. Read-only tools are available with the Read scope. Tools that create, update, or delete data require the Read + Write scope.| Area | What’s available |
|---|---|
| Search | Full-text search across projects, tasks, notes, calendar events, and links |
| Workspace meta | Who am I, list tags, list folders, list project types, workspace counts |
| Projects | Create, update, delete, set status, pin, unpin, list, get |
| Tasks | Create, update, delete, list, list subtasks, move between projects |
| Notes | Create, update, delete, list, plus full folder management |
| Calendar | Create, update, delete, list events |
| Time tracking | Read entries, create, update |
| Goals | Create, update, archive, list, get |
| Roadmap | Create, update, list, get |
| Notifications | List, mark as read |
| Analytics | Time summary, project health, productivity insights |
| Workflows | List, get, execute — your assistant can trigger your manual workflows |
Resources
The assistant can also read five built-in resources at any time:- Workspace summary — a snapshot of your whole workspace
- All projects — every project you have access to
- Today’s tasks — tasks due or scheduled for today
- This week’s calendar — upcoming events for the current week
- Single project with open tasks — detailed view of one project and its incomplete tasks
Prompt templates
The assistant can invoke four named prompts directly:| Prompt | What it does |
|---|---|
plan_my_week | Reviews your upcoming tasks and calendar, then drafts a prioritized weekly plan |
daily_standup | Reads yesterday’s completed tasks and today’s agenda, then writes a standup summary |
project_kickoff | Takes a new project and scaffolds milestones, first tasks, and suggested next steps |
task_retrospective | Summarizes a completed project’s tasks, timelines, and any patterns |
Walkthroughs
Daily standup in 30 seconds
You want a formatted standup summary to paste into Slack each morning.Ask your assistant to run the prompt
In Claude Desktop (or any connected client), type:
“Run the daily_standup prompt.”
The assistant reads your data
The assistant calls the
daily_standup prompt, which reads your tasks completed yesterday, your tasks due today, and your calendar for the day.plan_my_week the same way on Monday mornings, or project_kickoff immediately after creating a new project.
Triage a meeting transcript into tasks
You’ve just wrapped a meeting and have a transcript. You want action items to appear in Navis Ops without manual entry.Paste the transcript and give an instruction
In ChatGPT (or any connected client), paste the transcript and type:
“Read this transcript and create tasks for the action items, scoped to the Q3 launch project.”
The assistant finds your project
The assistant calls
list_projects to find the Q3 launch project’s ID.Tasks are created
It calls
create_task once per action item, setting the project ID, title, and due date from the transcript.Limitations
| What you’ll hit | Why |
|---|---|
| Two scopes only — read and write | There’s no per-tool access control. The Write scope grants all write tools. You can’t grant “create tasks but not delete projects.” |
| No rate limit enforced in-app | Rate limiting relies on Supabase platform limits. Heavy automated workloads can be throttled. |
| API keys never expire by default | Unless you set an expiration at creation time, the key works until you revoke it manually. |
| No live resource updates | Resources are fetched on demand. The assistant must re-read a resource to see data that changed since the last read. |
| You cannot read other users’ data | All tools are scoped to your authenticated account via row-level security. Cross-user reads are impossible. |
| API key is shown only once | If you lose it, revoke the key and create a new one. Navis Ops cannot retrieve it for you. |
FAQ
Which is better — API key or OAuth?
Which is better — API key or OAuth?
OAuth if your client supports it (Claude Desktop, Cursor). There’s no token to copy, store, or lose — you just approve in the browser. API keys work with every client and are faster to set up the first time.
Can I revoke an API key?
Can I revoke an API key?
Yes. Go to Settings → Connected Apps and click Revoke next to the key. The revocation is effective immediately — the next call from that client will fail.
Can I disconnect a Claude Desktop OAuth session?
Can I disconnect a Claude Desktop OAuth session?
Yes. Go to Settings → Connected Apps, find the client, and click Disconnect. Every active access and refresh token for that session is revoked immediately.
How do I see what my assistant has been doing?
How do I see what my assistant has been doing?
Each connection row in Settings → Connected Apps shows the last-used time. Per-call audit logs are not surfaced in-app yet.
My assistant says 'tool requires write scope.'
My assistant says 'tool requires write scope.'
The connection was made with the Read scope only. Create a new API key with Read + Write selected, or — for OAuth — disconnect and reconnect approving the write scope.
Will the assistant see other users' workspaces?
Will the assistant see other users' workspaces?
No. The bearer token scopes everything to your account via row-level security. The assistant cannot read or act on any other user’s data.
Can I limit which individual tools the assistant can call?
Can I limit which individual tools the assistant can call?
Not at the individual tool level. The only control is Read vs. Read + Write at the connection. If you want the assistant to only read data, use the Read scope.
See also
Workflows
Build automations your assistant can trigger using the Workflows execute tool.
Projects
Understand what the assistant can read and write in your projects.
MCP tools reference
Full reference for all 52 tools, their parameters, and required scopes.
API authentication
API key and OAuth details for developers integrating directly.