Skip to content
Sign in

For AI and agents

People,agent-readable.

HuMetric is a consent-first protocol for professional discovery. People choose what may be shared; participating systems must respect purpose, minimum-necessary use, contact preferences, correction and withdrawal. Agent access is MCP-based, authenticated and consent-logged. There is no public query API.

Public surface only

Protocols, standards, public documents and high-level technology names are shown here. Internal ranking, retrieval logic, data flows, models, infrastructure configuration and source code are intentionally excluded.

01. Protocol

Rules before access.

Shared expectations for every participating system

  1. 01

    Consent comes first

    A person chooses whether to participate, which parts of their professional context may be shared, and when that permission ends. Missing or private information must not be inferred.

  2. 02

    Purpose before access

    Use professional context only for a clear, legitimate purpose. Do not repurpose it for surveillance, manipulation or hidden eligibility decisions.

  3. 03

    Minimum necessary context

    Request only what is relevant to the stated purpose. A public profile is not permission to infer sensitive traits, reconstruct an identity or make unsupported claims.

  4. 04

    Respect the person’s intent

    Availability and contact preferences set the boundary for action. Agents must not create unwanted outreach, bypass a stated preference or pressure someone to engage.

  5. 05

    Accountable participation

    People should understand the purpose of an interaction, correct their public context and withdraw participation. Builders remain responsible for honoring those choices.

02. Access

Machine-readable access.

The public documents are live today. Agent access runs through MCP, authenticated and consent-logged.

Primary agent endpoint

POST /api/mcp

Live

MCP Streamable HTTP · OAuth 2.0 bearer access token

Available tools

discovery · profile · connect · inbox · team (workspace-gated)

Any signed-in member can search and fetch. Profile tools act only on your own profile. Members can start intros, and contact reveals (email, phone and profile links) are for paying members. Team tools require OAuth and an authorized team workspace.

Manage connected AI agents and their permissions

03. Connect

Connect an agent.

Add the server URL as a connector in your agent. It sends you to Humetric to sign in and approve access.

MCP server URL

https://humetric.net/api/mcp

Claude

  1. 01Open Customize → Connectors, then choose Add → Add custom connector.
  2. 02Enter Humetric as the name, paste the URL and click Continue, then click Add.
  3. 03Click Connect. The first time you ask Claude to use Humetric, it asks you to connect your account: approve access, then send your request again.

Claude Code

  1. 01Run this command in your terminal:claude mcp add --transport http humetric https://humetric.net/api/mcp
  2. 02In a Claude Code session, run /mcp, select humetric and choose Authenticate.
  3. 03Sign in to Humetric in the browser and approve access.

What happens next

  1. 01Your agent opens the Humetric sign-in page. Use your Humetric account.
  2. 02The consent screen lists the access the agent is asking for.
  3. 03You can approve less. Inbox access is optional and chosen separately.

Optional inbox permissions

inbox.read
List and read your own invitations and conversations.
inbox.write
Start intros, send invitations, reply and withdraw pending invitations.
invitations.decide
Accept or decline invitations.

These are separate from signing in. You can change them or disconnect the agent later at the link below. Manage connected AI agents and their permissions

What agents can do

Any signed-in member can use Discovery (search and fetch over profiles that people published and opened to AI-agent discovery), Profile tools for their own profile (including their own contact email, phone and profile links), Connect to start intros, and Inbox for their own conversations. Team tools need an authorized team workspace. Other people's contact (email, phone and profile links) is never returned except after an accepted intro that shares it, or a paid reveal.

The principle

People own their professional context. AI may help them be discovered, but never at the expense of consent, agency or dignity.

Read public protocol