Truestamp for AI agents

Connect your AI assistant to Truestamp and timestamp, search, and verify without leaving the conversation. It signs in as you, sees only what you can see, and asks your permission first.
$ claude mcp add --transport http truestamp https://www.truestamp.com/mcp

Or connect Claude Desktop, claude.ai, or another client

No API key to paste
Sign in once, in your browser
Scoped to your account
You approve what it can do
Three tools, not thirty

Building an integration?

Truestamp offers REST and GraphQL APIs for programmatic access from any language or platform.

Explore the API

Prefer a terminal?

The Truestamp CLI creates, downloads, and verifies proofs from a single static binary.

Explore the CLI

What your agent can do

Ordinary requests in plain language. Your assistant works out the rest and reports back.

Timestamp something, right from the chat

Hand it a file hash, a document, or a claim and it submits a timestamp for you, in the right team, without you switching windows.

See it before it happens

Ask for a dry run first. You get back the team it resolved to, the tidied-up tags, the visibility, and the exact fingerprint it would submit. Nothing is written.

Find your records

Search across everything you can see in every team you belong to. Filter by tag, state, visibility, or date, or pull up a single record by its ID.

Know when a record is ready to prove

Ask whether a timestamp is verifiable yet and get back the stage it is in, what it is waiting on, and roughly how long that takes.

Produce a proof, and check one

Generate a portable proof for a timestamp, a block, a beacon, or an entropy observation. Verify any proof someone hands you, including proofs you did not create.

Read the public ledger

Blocks, commitments to public blockchains, epochs, and the entropy observations that feed them. Public to your agent exactly as they are to anyone.

Ask it anything about Truestamp

Your assistant can search and read Truestamp's knowledge base on demand, so answers come from the source rather than from memory. Handy when you are wiring Truestamp into something else over REST or GraphQL.

Small things that save a round trip

Hash data with the algorithm you name, canonicalize JSON, read the time out of an ID, or work out what an unfamiliar identifier refers to.

Teams and notifications

Review your teams, their members, and pending invitations. Start a new team when you need somewhere to put new work, and set up webhook endpoints to be told when things happen.

What it will never do

Sensitive account and membership changes stay in the app, where you make them yourself.

  • Change your password, your email address, your role, or your billing.
  • Delete, redact, or un-redact a timestamp.
  • Transfer ownership of your records to someone else.
  • Invite people, remove people, or change anyone's role on a team.
  • Rename, reconfigure, or delete a team.
  • See anything you cannot see. Records belonging to teams you are not in are simply not there.

It runs as you, and only as you

The connection is built on the same standard your other connected apps use, so there is no secret to copy and no shared account.

No API key to paste

You sign in the way you always do. Your client points itself at Truestamp, Truestamp sends you to the usual sign-in page, you authenticate in your browser with your second factor, and the client receives a short-lived credential of its own. Nothing sensitive is ever copied into a config file.

You approve exactly what it can do

Before anything connects, a consent screen names the application asking, where it will send you back, and what it wants: permission to read your Truestamp data, and separately, permission to submit on your behalf. Reading is always required. Submitting is its own approval, and you can decline it and still use everything else.

Your permissions, unchanged

Every action runs as you. Same account, same teams, same role within each of those teams, same plan limits. There is no service account and no elevated mode. An agent connected to your account sees exactly what you see when you sign in, and nothing more.

Check the application name and the return address on the consent screen before you approve, the same way you would for any app asking to connect to your account.

Connect your client

Pick your client. Each one signs you in through your browser the first time you use it.

One command. No key, no header, no config file to edit by hand.

Add the server
$ claude mcp add --transport http truestamp https://www.truestamp.com/mcp

The first time you use it, Claude Code opens your browser to sign in and approve. Add --scope user to make it available in every project.

Claude Desktop, and any other client that reads an mcpServers block, take this configuration.

Client configuration
{
  "mcpServers": {
    "truestamp": {
      "type": "http",
      "url": "https://www.truestamp.com/mcp"
    }
  }
}

Drop this into your client's MCP configuration and restart it. It will prompt you to sign in the first time it connects.

Add Truestamp as a custom connector. Nothing to install.

  1. Open Settings, then Connectors.
  2. Choose Add custom connector.
  3. Paste the server address below and save.
Server address
https://www.truestamp.com/mcp

Sign-in and approval happen in the browser you are already using.

Some clients can only send a fixed API key header and cannot sign you in. Those cannot use this connector. Use the REST or GraphQL API with an API key instead.

Three tools, not thirty

Most connectors unload a long catalog of tool definitions into your session before your assistant does a single useful thing, and you pay for that catalog in every message that follows.

Truestamp advertises a handful of entry points and describes the rest only when they are actually needed. Several steps fold into one request instead of a chain of round trips. Your context window goes to your work, not to our menu.

Good to know

Built on an open standard

Truestamp speaks the Model Context Protocol over its Streamable HTTP transport. There is no Truestamp-specific protocol to learn and nothing proprietary to install, so any client that implements the standard can connect and you are never tied to one vendor's assistant.

2026-07-28 2025-06-18 2025-03-26

Supported revisions, newest first. Your client negotiates the newest one you both understand.

No team to pick first

The connection is not tied to a single team. Your assistant works across every team you belong to at once, and when it submits something it resolves which team the record belongs to rather than making you set a context up front. Ask for a dry run and it will tell you the team it picked before anything is written.

Stuck connecting, or want to know more about how any of this works? Visit support.