MCP Server

176 tools for AI agents — verify assets, manage approvals, and automate trust workflows

What is MCP?

The Model Context Protocol (MCP) lets AI assistants call tools directly. The proof.holdings MCP server wraps our entire REST API into 176 structured tools that any MCP-compatible agent can use — Claude, LangChain, CrewAI, OpenAI Agents SDK, Cursor, Windsurf, and more.

Two ways to connect

There is a hosted server and a local one. They expose the same tools; what differs is what you have to install and how you authenticate.

Hosted (remote)	Local (npm)
Install	nothing	`npx -y @proof-holdings/mcp-server`
Credential to start	none	none (`PROOF_API_KEY` is optional)
How an account is reached	browser sign-in (OAuth) from the client	`PROOF_API_KEY` in the environment
Transport	HTTP	stdio

Start with the hosted server unless you need a local process — it needs no install, and signing in happens in the client rather than by pasting a key into a config file.

Hosted server

bash

claude mcp add --transport http proof https://api.proof.holdings/mcp

The connection succeeds immediately and anonymously. The first tool that needs an account answers with a 401 carrying an OAuth challenge, and the client offers to authenticate — in Claude Code that is /mcp → Authenticate. You sign in and approve the request in a browser; the client stores the result and sends it from then on.

Approving that request creates a scoped API key on your account, named after the client. Revoking access is deleting that key in the dashboard.

Local server

bash

npx -y @proof-holdings/mcp-server

It starts with no configuration. Supplying PROOF_API_KEY is what gives it an account from the first call.

Variable	Required	Default	Description
`PROOF_API_KEY`	No	—	Your API key (`pk_live_` or `pk_test_`). Without it the server still starts.
`PROOF_BASE_URL`	No	`https://api.proof.holdings`	API base URL

What works without an account

The server starts and lists all 176 tools whether or not it holds a credential — an agent can always see what is available. Ten tools then work with no account at all, in four groups:

Checking somebody else's proof — validate_proof verifies a signed token against our published keys, and list_revoked_proofs reads the revocation list.
Checking somebody else's delegation — verify_delegation answers whether one artifact is authorized by the domain it claims, and verify_delegations re-checks a set of them. This is the half an agent reaches for before installing something.
Signing in — start_login and wait_for_login run a browser or messenger sign-in and return a session.
Creating an account — create_account, send_account_email and wait_for_account_creation bootstrap a new account from inside the agent.

Plus render_auth_link, which formats a sign-in link locally and calls nothing.

The verification tools are open on purpose: the party checking a proof or a delegation is usually not our customer, and a check that required an account would not get run.

Everything else needs an account. Those tools do not fail obscurely: each says an account is needed and how to get one.

Signing in with start_login gives a session, and a session reaches 54 of the tools — the ones that act on your own account through the dashboard's own authentication. The remaining tools need an API key (or the hosted server's OAuth flow, which mints one).

Client setup

Claude Code CLI

bash

# hosted
claude mcp add --transport http proof https://api.proof.holdings/mcp

# local
claude mcp add proof-holdings -- npx -y @proof-holdings/mcp-server \
  --env PROOF_API_KEY=pk_live_YOUR_API_KEY

Claude Desktop, Cursor, Windsurf

Every client below takes the same two shapes. The hosted one needs no key:

json

{
  "mcpServers": {
    "proof": {
      "type": "http",
      "url": "https://api.proof.holdings/mcp"
    }
  }
}

The local one runs the package, and PROOF_API_KEY is optional — without it the server still starts and the keyless tools above work:

json

{
  "mcpServers": {
    "proof-holdings": {
      "command": "npx",
      "args": ["-y", "@proof-holdings/mcp-server"],
      "env": {
        "PROOF_API_KEY": "pk_live_YOUR_API_KEY"
      }
    }
  }
}

GET /api/v1/mcp/connect serves both fragments live, so an agent can read the current ones rather than copying from here.

Config file locations:

Client	Config file
Claude Desktop (macOS)	`~/Library/Application Support/Claude/claude_desktop_config.json`
Claude Desktop (Windows)	`%APPDATA%\Claude\claude_desktop_config.json`
Cursor	`.cursor/mcp.json` in your project
Windsurf	`~/.codeium/windsurf/mcp_config.json`

Tool Overview (176 tools)

Group totals below are exact. The tables name the tools you are most likely to reach for rather than all 176 — an MCP client's own tools/list is the complete, current list, and it is the one the server answers from.

Verifications & requests (28 tools)

Asset control challenges, multi-channel phone verification, and the multi-asset B2B request flow.

Tool	Description
`create_verification`	Create a new verification challenge
`get_verification`	Get verification status
`submit_verification_code`	Submit a verification code
`trigger_verification`	Trigger a verification check
`wait_for_verification`	Poll until verification completes
`create_multi_channel_verification`	Challenge one phone across up to three channels, first completion wins
`create_verification_request`	Create a multi-asset request
`get_request_by_reference`	Get a request by reference ID
`share_request_email`	Email a request to its recipient

Domains & DNS (23 tools)

Domain control proofs, DNS provider credentials, and the email-based domain flow.

Tool	Description
`add_domain`	Add a domain to verify
`verify_domain`	Check the challenge record and mint the proof
`connect_cloudflare`	Connect Cloudflare so records are written for you
`verify_domain_with_credentials`	Prove control using stored provider credentials
`setup_domain_email`	Set up sending from the domain
`list_dns_credentials`	List stored DNS provider credentials

Account, settings & billing (37 tools)

Everything about your own account: usage, keys, contact identifiers, 2FA, and the agent account bootstrap.

Tool	Description
`get_platform_summary`	One-call snapshot of the account
`get_usage`	Quota and usage for the period
`search`	Search across the account
`create_account`	Bootstrap a new account from inside an agent
`create_api_key`	Create a scoped API key
`list_assets`	List verified assets and their proof handles
`start_2fa`	Begin a 2FA step-up

Human-in-the-loop approvals, their end-to-end encrypted key material, and the consent layer underneath them.

Tool	Description
`create_hitl`	Create a human-approval configuration
`create_confirmation`	Send an approval request to a person
`wait_for_confirmation`	Poll until a person approves or denies
`get_confirmation_approval_link`	Get a fresh approval link
`create_authorization`	Ask a person to consent to being contacted
`revoke_authorization`	Withdraw a consent

Circles & Proof of Me (20 tools)

The people primitive: circles of trusted contacts, their channels, and cross-channel identity challenges.

Tool	Description
`create_circle`	Create a circle
`add_circle_member`	Add a trusted contact
`invite_circle_member`	Send a single-use enrollment link
`add_circle_member_channel`	Add a channel to a member
`create_identity_challenge`	Run a cross-channel identity check

Public profiles (16 tools)

Public verification pages, the proofs they display, and their templates.

Tool	Description
`create_profile`	Create a public profile
`claim_username`	Claim a public username
`update_public_proofs`	Choose which proofs a profile shows
`preview_profile_template`	Preview a profile template

Templates & webhooks (11 tools)

Message templates for outbound channels, and the webhook delivery log.

Tool	Description
`list_templates`	List message templates
`update_template`	Change a template
`render_template`	Render a template with variables
`list_webhook_deliveries`	List webhook deliveries
`retry_webhook_delivery`	Retry a failed delivery

Delegations & proofs (10 tools)

Proof of Delegation — a proven domain authorizing an artifact — plus the proof lifecycle and the public verifier surface.

Tool	Description
`create_delegation`	Authorize an artifact from a domain you have proven
`list_delegations`	List your delegations
`revoke_delegation`	Revoke a delegation
`verify_delegation`	Check whether an artifact is authorized by the domain it claims
`validate_proof`	Verify a signed proof token (no account needed)
`get_proof_status`	Read a proof's status by its public handle
`list_revoked_proofs`	Read the revocation list (no account needed)

verify_delegation is the one to reach for when an agent is deciding whether to trust a server or package. It answers about an artifact identity you resolved independently — a package you are installing, an endpoint you are calling — never one read out of the artifact's own files.

Browser and messenger sign-in, and the sessions it produces.

Tool	Description
`start_login`	Begin a sign-in (no account needed)
`wait_for_login`	Poll until sign-in completes
`get_current_user`	Who the current session belongs to
`list_auth_sessions`	List active sessions
`render_auth_link`	Render a sign-in link for the user to open

Example: Verify a phone number

text

Agent: I need to verify that +1-555-0123 belongs to this user.

→ create_verification({ type: "phone", identifier: "+15550123", channel: "sms" })
← { id: "ver_abc123", status: "pending", expires_at: "..." }

→ wait_for_verification({ id: "ver_abc123" })
← { status: "verified", proof_token: "eyJhbGc..." }

Example: Request human approval

text

Agent: This wire transfer needs a human to approve it.

→ create_confirmation({ hitl_id: "hitl_abc", message: "<ciphertext>" })
← { id: "conf_xyz", status: "pending", timeout_at: "..." }

→ wait_for_confirmation({ id: "conf_xyz" })
← { status: "resolved", response: { action: "approve" }, proof_token: "eyJhbGc..." }

Example: Check whether a server may act for a domain

text

Agent: This MCP server says it works on behalf of example.com. Is that true?

→ verify_delegation({ delegate: "pkg:npm/their-package", principal: "example.com" })
← { verified: true, scope: ["send-email"], expires_at: "..." }

Test mode

A pk_test_* key puts every tool in test mode: verifications complete without sending real messages, and test data is kept apart from production data. test_verify completes a test verification directly.

Resources

API Reference — the REST API these tools wrap
Proof of Delegation — the delegation module in full
SDKs — JavaScript, Python, Go and PHP clients
Core Primitives — the mental model in one page

API Reference SDKs Integrations MCP Server Ownership

MCP Server

176 tools for AI agents — verify assets, manage approvals, and automate trust workflows

What is MCP?

Two ways to connect

There is a hosted server and a local one. They expose the same tools; what differs is what you have to install and how you authenticate.

Hosted (remote)	Local (npm)
Install	nothing	`npx -y @proof-holdings/mcp-server`
Credential to start	none	none (`PROOF_API_KEY` is optional)
How an account is reached	browser sign-in (OAuth) from the client	`PROOF_API_KEY` in the environment
Transport	HTTP	stdio

Start with the hosted server unless you need a local process — it needs no install, and signing in happens in the client rather than by pasting a key into a config file.

Hosted server

bash

claude mcp add --transport http proof https://api.proof.holdings/mcp

Approving that request creates a scoped API key on your account, named after the client. Revoking access is deleting that key in the dashboard.

Local server

bash

npx -y @proof-holdings/mcp-server

It starts with no configuration. Supplying PROOF_API_KEY is what gives it an account from the first call.

Variable	Required	Default	Description
`PROOF_API_KEY`	No	—	Your API key (`pk_live_` or `pk_test_`). Without it the server still starts.
`PROOF_BASE_URL`	No	`https://api.proof.holdings`	API base URL

What works without an account

The server starts and lists all 176 tools whether or not it holds a credential — an agent can always see what is available. Ten tools then work with no account at all, in four groups:

Checking somebody else's proof — validate_proof verifies a signed token against our published keys, and list_revoked_proofs reads the revocation list.
Checking somebody else's delegation — verify_delegation answers whether one artifact is authorized by the domain it claims, and verify_delegations re-checks a set of them. This is the half an agent reaches for before installing something.
Signing in — start_login and wait_for_login run a browser or messenger sign-in and return a session.
Creating an account — create_account, send_account_email and wait_for_account_creation bootstrap a new account from inside the agent.

Plus render_auth_link, which formats a sign-in link locally and calls nothing.

The verification tools are open on purpose: the party checking a proof or a delegation is usually not our customer, and a check that required an account would not get run.

Everything else needs an account. Those tools do not fail obscurely: each says an account is needed and how to get one.

Client setup

Claude Code CLI

bash

# hosted
claude mcp add --transport http proof https://api.proof.holdings/mcp

# local
claude mcp add proof-holdings -- npx -y @proof-holdings/mcp-server \
  --env PROOF_API_KEY=pk_live_YOUR_API_KEY

Claude Desktop, Cursor, Windsurf

Every client below takes the same two shapes. The hosted one needs no key:

json

{
  "mcpServers": {
    "proof": {
      "type": "http",
      "url": "https://api.proof.holdings/mcp"
    }
  }
}

The local one runs the package, and PROOF_API_KEY is optional — without it the server still starts and the keyless tools above work:

json

{
  "mcpServers": {
    "proof-holdings": {
      "command": "npx",
      "args": ["-y", "@proof-holdings/mcp-server"],
      "env": {
        "PROOF_API_KEY": "pk_live_YOUR_API_KEY"
      }
    }
  }
}

GET /api/v1/mcp/connect serves both fragments live, so an agent can read the current ones rather than copying from here.

Config file locations:

Client	Config file
Claude Desktop (macOS)	`~/Library/Application Support/Claude/claude_desktop_config.json`
Claude Desktop (Windows)	`%APPDATA%\Claude\claude_desktop_config.json`
Cursor	`.cursor/mcp.json` in your project
Windsurf	`~/.codeium/windsurf/mcp_config.json`

Tool Overview (176 tools)

Verifications & requests (28 tools)

Asset control challenges, multi-channel phone verification, and the multi-asset B2B request flow.

Tool	Description
`create_verification`	Create a new verification challenge
`get_verification`	Get verification status
`submit_verification_code`	Submit a verification code
`trigger_verification`	Trigger a verification check
`wait_for_verification`	Poll until verification completes
`create_multi_channel_verification`	Challenge one phone across up to three channels, first completion wins
`create_verification_request`	Create a multi-asset request
`get_request_by_reference`	Get a request by reference ID
`share_request_email`	Email a request to its recipient

Domains & DNS (23 tools)

Domain control proofs, DNS provider credentials, and the email-based domain flow.

Tool	Description
`add_domain`	Add a domain to verify
`verify_domain`	Check the challenge record and mint the proof
`connect_cloudflare`	Connect Cloudflare so records are written for you
`verify_domain_with_credentials`	Prove control using stored provider credentials
`setup_domain_email`	Set up sending from the domain
`list_dns_credentials`	List stored DNS provider credentials

Account, settings & billing (37 tools)

Everything about your own account: usage, keys, contact identifiers, 2FA, and the agent account bootstrap.

Tool	Description
`get_platform_summary`	One-call snapshot of the account
`get_usage`	Quota and usage for the period
`search`	Search across the account
`create_account`	Bootstrap a new account from inside an agent
`create_api_key`	Create a scoped API key
`list_assets`	List verified assets and their proof handles
`start_2fa`	Begin a 2FA step-up

Human-in-the-loop approvals, their end-to-end encrypted key material, and the consent layer underneath them.

Tool	Description
`create_hitl`	Create a human-approval configuration
`create_confirmation`	Send an approval request to a person
`wait_for_confirmation`	Poll until a person approves or denies
`get_confirmation_approval_link`	Get a fresh approval link
`create_authorization`	Ask a person to consent to being contacted
`revoke_authorization`	Withdraw a consent

Circles & Proof of Me (20 tools)

The people primitive: circles of trusted contacts, their channels, and cross-channel identity challenges.

Tool	Description
`create_circle`	Create a circle
`add_circle_member`	Add a trusted contact
`invite_circle_member`	Send a single-use enrollment link
`add_circle_member_channel`	Add a channel to a member
`create_identity_challenge`	Run a cross-channel identity check

Public profiles (16 tools)

Public verification pages, the proofs they display, and their templates.

Tool	Description
`create_profile`	Create a public profile
`claim_username`	Claim a public username
`update_public_proofs`	Choose which proofs a profile shows
`preview_profile_template`	Preview a profile template

Templates & webhooks (11 tools)

Message templates for outbound channels, and the webhook delivery log.

Tool	Description
`list_templates`	List message templates
`update_template`	Change a template
`render_template`	Render a template with variables
`list_webhook_deliveries`	List webhook deliveries
`retry_webhook_delivery`	Retry a failed delivery

Delegations & proofs (10 tools)

Proof of Delegation — a proven domain authorizing an artifact — plus the proof lifecycle and the public verifier surface.

Tool	Description
`create_delegation`	Authorize an artifact from a domain you have proven
`list_delegations`	List your delegations
`revoke_delegation`	Revoke a delegation
`verify_delegation`	Check whether an artifact is authorized by the domain it claims
`validate_proof`	Verify a signed proof token (no account needed)
`get_proof_status`	Read a proof's status by its public handle
`list_revoked_proofs`	Read the revocation list (no account needed)

Browser and messenger sign-in, and the sessions it produces.

Tool	Description
`start_login`	Begin a sign-in (no account needed)
`wait_for_login`	Poll until sign-in completes
`get_current_user`	Who the current session belongs to
`list_auth_sessions`	List active sessions
`render_auth_link`	Render a sign-in link for the user to open

Example: Verify a phone number

text

Agent: I need to verify that +1-555-0123 belongs to this user.

→ create_verification({ type: "phone", identifier: "+15550123", channel: "sms" })
← { id: "ver_abc123", status: "pending", expires_at: "..." }

→ wait_for_verification({ id: "ver_abc123" })
← { status: "verified", proof_token: "eyJhbGc..." }

Example: Request human approval

text

Agent: This wire transfer needs a human to approve it.

→ create_confirmation({ hitl_id: "hitl_abc", message: "<ciphertext>" })
← { id: "conf_xyz", status: "pending", timeout_at: "..." }

→ wait_for_confirmation({ id: "conf_xyz" })
← { status: "resolved", response: { action: "approve" }, proof_token: "eyJhbGc..." }

Example: Check whether a server may act for a domain

text

Agent: This MCP server says it works on behalf of example.com. Is that true?

→ verify_delegation({ delegate: "pkg:npm/their-package", principal: "example.com" })
← { verified: true, scope: ["send-email"], expires_at: "..." }

Test mode

Resources

API Reference — the REST API these tools wrap
Proof of Delegation — the delegation module in full
SDKs — JavaScript, Python, Go and PHP clients
Core Primitives — the mental model in one page

API Reference SDKs Integrations MCP Server Ownership

MCP Server

What is MCP?

Two ways to connect

Hosted server

Local server

What works without an account

Client setup

Claude Code CLI

Claude Desktop, Cursor, Windsurf

Tool Overview (176 tools)

Verifications & requests (28 tools)

Domains & DNS (23 tools)

Account, settings & billing (37 tools)

HITL approvals & consent (22 tools)

Circles & Proof of Me (20 tools)

Public profiles (16 tools)

Templates & webhooks (11 tools)

Delegations & proofs (10 tools)

Sign-in & sessions (9 tools)

Example: Verify a phone number

Example: Request human approval

Example: Check whether a server may act for a domain

Test mode

Resources

Related

MCP Server

What is MCP?

Two ways to connect

Hosted server

Local server

What works without an account

Client setup

Claude Code CLI

Claude Desktop, Cursor, Windsurf

Tool Overview (176 tools)

Verifications & requests (28 tools)

Domains & DNS (23 tools)

Account, settings & billing (37 tools)

HITL approvals & consent (22 tools)

Circles & Proof of Me (20 tools)

Public profiles (16 tools)

Templates & webhooks (11 tools)

Delegations & proofs (10 tools)

Sign-in & sessions (9 tools)

Example: Verify a phone number

Example: Request human approval

Example: Check whether a server may act for a domain

Test mode

Resources

Related