Integrations

MCP and HTTP

Connect Trustvice to the tools you already use.

One MCP server and one Agentic API, behind the same gateway. Connect an AI client or your own backend to the Trustvice 360 action catalogue.

Browse the directory Connect an MCP client

16 Trustvice 360 suites behind one gateway
MCP and the Agentic API
Public keys, offline verification

Where calls come from

AI clients over MCP
Your code over the Agentic API

One gateway

Authenticates the credential, checks the action's permissions, applies rate limits and writes the audit entry. The same checks for every transport.

One response envelope

ok and data on success, or an error with a machine-readable code, on every transport.

Anyone can verify

Signed records verify against Trustvice's published public keys. No account needed.

How to read this directory

Each entry says what exists today. Where something is not published yet, we say so and show the way that works now.

Hosted service: Runs on Trustvice. Point your client at the address and authenticate.
Published: A package or file you can download from a public registry or URL today.
Not published yet: The source is complete but not released. Use the alternative shown on the card.
Manifest only: A description file is published for hosts that read it. We do not document install steps.

Start here

Quick start

Three ways to start

Connect an assistant over MCP, call the Agentic API from your code, or verify signed records yourself.

MCP server

Hosted service

Gives Claude, Cursor, ChatGPT and any client that speaks streamable HTTP the Trustvice 360 actions as tools, projected from the Agentic API catalogue.

Stateless JSON-RPC over HTTP POST, protocol version 2025-06-18
Workspace key, personal token or OAuth with PKCE
The same permissions and confirmations as the Agentic API

bash

curl -sS https://mcp.trustvice.com/mcp \
  -H "Authorization: Bearer $TRUSTVICE_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{ "jsonrpc": "2.0", "id": 2, "method": "tools/list" }'

MCP setup

Agentic API

Hosted service

One endpoint for every declared action: POST /api/v1/actions/<id>, with a typed input and a uniform response envelope.

The public catalogue is served at GET /api/v1/actions, with no credential
Every action declares a safety level: read, write, destructive or billing_impacting
Destructive and billing actions ask for a short-lived confirmation token first

HTTP

POST /api/v1/actions/<action-id>
Authorization: Bearer <oauth-access-token>
X-Workspace-Id: ws_…
Content-Type: application/json

{ …input fields… }

→ { "ok": true, "data": { …action output… } }

Agentic actions

Verifier library (not published yet)

Not published yet

Verifies signed receipts, Notarian evidence bundles, TOPS seals and audit-chain exports with WebCrypto alone, fully offline. Until it is on the npm registry, use the verify page.

Zero dependencies
Returns bad_signature, unknown_key or malformed_envelope when a check fails
A valid signature proves integrity and origin, not that a claim inside is true

verify.ts

import { verifyArtifactWithKeys } from "@trustvice/verify";

const jwks = await (await fetch("https://api.trustvice.com/.well-known/trustvice-keys.json")).json();
const verdict = await verifyArtifactWithKeys(receipt, jwks);
if (!verdict.ok) throw new Error(verdict.reason);

Open the verify page

Every way to connect

Filter by category. Each card gives the setup in a few lines and links to the page with the full steps.

Showing 21 of 21

Claude Code

AI clientsHosted service

Adds Trustvice as a remote HTTP MCP server; sign in with OAuth through the /mcp command.

bash

claude mcp add --transport http trustvice https://mcp.trustvice.com/mcp

Read the docs

Claude on the web and desktop

AI clientsHosted service

Connects as a custom connector over OAuth with PKCE.

Claude

Settings → Connectors → Add custom connector
Name:   Trustvice
URL:    https://mcp.trustvice.com/mcp
Then:   Connect, sign in with your Trustvice account

Read the docs

Cursor

AI clientsHosted service

Reads the server from mcp.json for all projects or one, then signs in with OAuth or sends a key from your environment.

~/.cursor/mcp.json

{
  "mcpServers": {
    "trustvice": {
      "url": "https://mcp.trustvice.com/mcp"
    }
  }
}

Read the docs

ChatGPT

AI clientsHosted service

Connects as a custom connector in developer mode with OAuth, where your plan and workspace admin allow it.

ChatGPT

Settings → Connectors → Advanced → Developer mode
Create → Name: Trustvice
MCP server URL: https://mcp.trustvice.com/mcp
Authentication: OAuth → Create → Connect

Read the docs

OpenAI Responses API

AI clientsHosted service

Uses Trustvice as a remote MCP tool from scripts; keep require_approval set to always so a person approves each action.

json

{
  "type": "mcp",
  "server_label": "trustvice",
  "server_url": "https://mcp.trustvice.com/mcp",
  "authorization": "<OAuth access token>",
  "require_approval": "always"
}

Read the docs

VS Code and other MCP clients

AI clientsHosted service

Any client that supports remote HTTP servers with custom headers can connect. Send the X-Workspace-Id header with your credential.

MCP

type     http (streamable HTTP)
url      https://mcp.trustvice.com/mcp
header   Authorization: Bearer <credential>
header   X-Workspace-Id: ws_…

Read the docs

@trustvice/verify (not published yet)

SDKs and librariesNot published yet

Verifies a signed record against a key set you pass in, with no network call. Not on the npm registry yet; the verify page works today.

import { verifyArtifactWithKeys } from "@trustvice/verify";

const verdict = await verifyArtifactWithKeys(receipt, jwks);
console.log(verdict.ok ? verdict.kind : verdict.reason);

Open the verify page

Agentic API

APIs and specsHosted service

Every declared action as one POST over plain HTTP, the same catalogue the MCP server projects as tools.

bash

curl -sS -X POST https://api.trustvice.com/v1/actions/trustactor.runs.list \
  -H "Authorization: Bearer $TRUSTVICE_TOKEN" \
  -H "X-Workspace-Id: ws_…" \
  -H "Content-Type: application/json" -d '{}'

Read the docs

Action catalogue

APIs and specsHosted service

The public, self-describing list of registered actions, served anonymously.

bash

curl -s https://api.trustvice.com/v1/actions

Read the docs

OpenAPI 3.1 document

APIs and specsPublished

Describes the action API, its schemas and security schemes, at /openapi.json and the well-known path.

bash

curl -s https://360.trustvice.com/openapi.json -o trustvice-openapi.json
# also at /.well-known/openapi.json

Read the docs

Public signing keys

APIs and specsPublished

The Ed25519 public keys that sign receipts, as a JSON key set.

bash

curl -sSL https://360.trustvice.com/.well-known/receipt-keys.json

Connect Trustvice to the tools you already use.

How to read this directory

Three ways to start

MCP server

Agentic API

Verifier library (not published yet)

Every way to connect

Claude Code

Claude on the web and desktop

Cursor

ChatGPT

OpenAI Responses API

VS Code and other MCP clients

@trustvice/verify (not published yet)

Agentic API

Action catalogue

OpenAPI 3.1 document

Public signing keys

llms.txt

AGENTS.md

MCP server manifests

Agent card

OAuth metadata

ChatGPT app manifest

Gemini extension manifest

Perplexity connector manifest

Claude connector manifest

NLWeb

Reachable from any agent that speaks MCP or HTTP

No integration for your stack? Call the API.

Agentic API

OpenAPI and the action API

Verification

What every verifier does

The same rules for every integration

One gateway

Least privilege

Confirmation for risky actions

Keys stay server side

Missing an integration?

Declare the rules. Have every run checked.

Connect Trustvice to the tools you already use.

How to read this directory

Three ways to start

MCP server

Agentic API

Verifier library (not published yet)

Every way to connect

Claude Code

Claude on the web and desktop

Cursor

ChatGPT

OpenAI Responses API

VS Code and other MCP clients

@trustvice/verify (not published yet)

Agentic API

Action catalogue

OpenAPI 3.1 document

Public signing keys

llms.txt

AGENTS.md

MCP server manifests

Agent card

OAuth metadata

ChatGPT app manifest

Gemini extension manifest

Perplexity connector manifest

Claude connector manifest

NLWeb

Reachable from any agent that speaks MCP or HTTP

No integration for your stack? Call the API.

Agentic API

OpenAPI and the action API

Verification

What every verifier does

The same rules for every integration

One gateway

Least privilege

Confirmation for risky actions

Keys stay server side