Tend — MCP connector
Read-only access to your own Tend record — goals, contacts, commitments and interaction history — so your assistant can answer from your network instead of guessing.
https://www.tendladder.com/api/mcp
Streamable HTTP · OAuth 2.1 · read-only · requires a Tend account (the free plan includes the connector).
This repository is the connector's documentation and its registry manifest. The server itself is hosted — there is nothing here to install or run.
What it is for
Tend is goal-coupled relationship management: your network organized around what you are trying to achieve, rather than a contact list on a reminder timer.
The connector exists because the useful question is usually asked somewhere Tend has no app — in the middle of a conversation with your assistant, on a phone, in a thread about a job you are trying to fill. It answers from your own record, and it never spends your Tend allowance: the reasoning runs on your assistant's plan, not ours.
Connecting
- In Claude, add a custom connector and paste
https://www.tendladder.com/api/mcp. - Click Allow on the consent screen — it appears on
tendladder.com, not a third-party login. - That is the whole setup. No token to paste, no config file to edit.
Verified on Claude desktop, web and mobile. Other MCP clients speak the same protocol and are expected to work, but only Claude is tested — ChatGPT requires Developer mode, and Gemini currently has no consumer path for remote MCP servers.
Every connected client is listed in Tend's settings with its own Revoke.
What to ask
The connector tells your assistant the questions it answers best, so asking "what can I do with Tend?" returns a useful list on the first try. These are the six it names:
| Ask | What it does |
|---|---|
| What do I owe people this week? | Your open commitments, the ones you promised |
| What am I waiting on? | The other direction — what others owe you |
| Who should I contact today, and why that person? | Ranked, with the reason attached |
| Which of my goals is stalling? | The goal that has stopped moving, and why |
| Brief me on Dana before I speak to her. | Everything on one person, in reading order |
| Here is a job spec — who in my network is relevant? | Paste anything; it matches against your network |
That last one is the one Tend's own app cannot do, because it cannot see what is in front of you and your assistant can.
Tools
Seven, all read-only.
| Tool | Returns |
|---|---|
list_goals | Your goals and their state |
get_goal_next_moves | The next moves on a goal, with the audit's own reasoning and its age |
list_commitments | Open commitments, each carrying a direction — mine or theirs |
search_contacts | Contacts matching a query, ranked |
get_contact | One person in full |
search_activities | Interaction history |
list_reminders | What is scheduled |
Rows carry deep links back into Tend, so anything your assistant creates elsewhere stays traceable to why it mattered.
Read-only is a hard boundary, not a first version's limitation. No tool creates, edits, sends or deletes. The worst outcome of a confused agent is a wrong answer, never a changed record — which is the property that made shipping this early defensible in the first place.
Privacy
- Your data is single-tenant and encrypted. It is never sold and never used to train models.
- Tenancy is resolved from your OAuth session, never from anything the model says — the server cannot be talked into reading someone else's record.
- Full policy: https://www.tendladder.com/privacy
Registry
Published to the Official MCP Registry
as com.tendladder/tend. The manifest is server.json.
Support
Questions, or something behaving oddly: open an issue here, or use the contact route at https://www.tendladder.com/.
License
The documentation and manifest in this repository are MIT licensed (see LICENSE). The Tend service itself is proprietary and is not distributed here.