> ## Documentation Index
> Fetch the complete documentation index at: https://docs.grantex.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Agent Framework Integrations

> Choose Grantex integrations for OpenAI Agents, Anthropic, LangChain, CrewAI, Google ADK, AutoGen, Vercel AI, MCP, Express, FastAPI, Strands, and A2A.

## Choose an AI agent integration

<CardGroup cols={2}>
  <Card title="Service Provider Adapters" icon="puzzle-piece" href="/integrations/adapters">
    Pre-built integrations for 11 services: Google Calendar, Gmail, Drive, Stripe, Slack, GitHub, Notion, HubSpot, Salesforce, Linear, Jira.
  </Card>

  <Card title="Credential Vault" icon="vault" href="/api-reference/vault/store">
    Encrypted per-user credential store. Agents exchange grant tokens for upstream service credentials.
  </Card>

  <Card title="MCP Auth Server" icon="lock" href="/integrations/mcp-auth">
    OAuth 2.1 + PKCE endpoint package for single-process evaluation; review the version 2.0.2 consent, code-handoff, state, and revocation limitations.
  </Card>

  <Card title="Grantex Gateway" icon="shield-halved" href="/integrations/gateway">
    Zero-code reverse-proxy that enforces grant tokens via YAML config.
  </Card>

  <Card title="Express.js" icon="server" href="/integrations/express">
    Grant token verification and scope-based authorization middleware.
  </Card>

  <Card title="FastAPI" icon="bolt-lightning" href="/integrations/fastapi">
    Dependency injection for grant token verification and scope checks.
  </Card>

  <Card title="MCP Server" icon="plug" href="/integrations/mcp">
    17 tools for Claude Desktop, Cursor, and Windsurf.
  </Card>

  <Card title="LangChain" icon="link-horizontal" href="/integrations/langchain">
    Scope-enforced tools and audit callbacks.
  </Card>

  <Card title="Vercel AI SDK" icon="bolt" href="/integrations/vercel-ai">
    Zod-schema tools with scope checks at construction time.
  </Card>

  <Card title="AutoGen / OpenAI" icon="robot" href="/integrations/autogen">
    Function calling with scope enforcement and registry.
  </Card>

  <Card title="CrewAI" icon="users" href="/integrations/crewai">
    Python tools with scope enforcement and audit logging.
  </Card>

  <Card title="OpenAI Agents SDK" icon="microchip" href="/integrations/openai-agents">
    FunctionTool wrapper with JWKS-backed token verification and scope checks.
  </Card>

  <Card title="Google ADK" icon="brain" href="/integrations/google-adk">
    Plain function tools for Google Agent Development Kit.
  </Card>

  <Card title="Strands Agents SDK" icon="microchip" href="/integrations/strands">
    TypeScript and Python Strands tools with Grantex scope enforcement.
  </Card>

  <Card title="Anthropic SDK" icon="message-bot" href="/integrations/anthropic">
    Scope-enforced tool use and audit logging for Claude models.
  </Card>

  <Card title="x402 Payment Protocol" icon="credit-card" href="/integrations/x402">
    Agent spend authorization for x402 payment flows — GDTs for USDC on Base L2.
  </Card>

  <Card title="CLI" icon="terminal" href="/integrations/cli">
    Manage agents, grants, and audit from your terminal.
  </Card>

  <Card title="A2A Bridge (TypeScript)" icon="arrows-left-right" href="/integrations/a2a">
    Inject Grantex grant tokens into Google A2A agent-to-agent communication.
  </Card>

  <Card title="A2A Bridge (Python)" icon="arrows-left-right" href="/integrations/a2a-py">
    Python client/server for Grantex-authenticated A2A task exchange.
  </Card>

  <Card title="Gemma 4 (On-Device)" icon="microchip" href="/integrations/gemma">
    Verify signed grant claims against a provisioned key bundle during offline operation; refresh keys, token validity, and revocation data through a trusted synchronization path.
  </Card>

  <Card title="DPDP / GDPR Technical Controls" icon="shield-check" href="/integrations/dpdp">
    Consent records, purpose-limitation fields, grievance workflows, and audit exports can support DPDP Act 2023 and EU GDPR programs; they are technical tooling, not legal compliance or certification.
  </Card>

  <Card title="Trust Registry" icon="building-columns" href="/integrations/registry">
    Public directory of registered AI agent publishers with DID profiles, public keys, and DNS ownership verification. Stored badge fields are metadata, not a compliance certification program.
  </Card>

  <Card title="Scope Enforcement" icon="shield-check" href="/guides/scope-enforcement">
    Enforce tool-level permissions on any connector — define your own manifests or use 53 pre-built ones. One-line enforcement via enforce().
  </Card>

  <Card title="Conformance Suite" icon="check" href="/integrations/conformance">
    Validate any Grantex server against the spec.
  </Card>

  <Card title="Event Destinations" icon="satellite-dish" href="/guides/event-streaming">
    Forward Grantex events to Datadog, Splunk, S3, BigQuery, and Kafka.
  </Card>

  <Card title="Terraform Provider" icon="cubes" href="/guides/terraform">
    Manage Grantex resources as infrastructure with the official Terraform provider.
  </Card>
</CardGroup>

## At a Glance

| Framework                   | Package                      | Install                                                   | Language                           |
| --------------------------- | ---------------------------- | --------------------------------------------------------- | ---------------------------------- |
| Adapters                    | `@grantex/adapters`          | `npm install @grantex/adapters`                           | TypeScript                         |
| MCP Auth Server             | `@grantex/mcp-auth@2.0.2`    | `npm install @grantex/mcp-auth@2.0.2 @grantex/sdk@0.3.13` | TypeScript; evaluation only        |
| Gateway                     | `@grantex/gateway`           | `npm install @grantex/gateway`                            | TypeScript                         |
| Express.js                  | `@grantex/express`           | `npm install @grantex/express`                            | TypeScript                         |
| FastAPI                     | `grantex-fastapi`            | `pip install grantex-fastapi`                             | Python                             |
| MCP                         | `@grantex/mcp`               | `npm install -g @grantex/mcp`                             | TypeScript                         |
| LangChain                   | `@grantex/langchain`         | `npm install @grantex/langchain`                          | TypeScript                         |
| Vercel AI SDK               | `@grantex/vercel-ai`         | `npm install @grantex/vercel-ai`                          | TypeScript                         |
| AutoGen / OpenAI            | `@grantex/autogen`           | `npm install @grantex/autogen`                            | TypeScript                         |
| CrewAI                      | `grantex-crewai`             | `pip install grantex-crewai`                              | Python                             |
| OpenAI Agents SDK           | `grantex-openai-agents`      | `pip install grantex-openai-agents`                       | Python                             |
| Google ADK                  | `grantex-adk`                | `pip install grantex-adk`                                 | Python                             |
| Strands Agents SDK (TS)     | `@grantex/strands`           | `npm install @grantex/strands`                            | TypeScript                         |
| Strands Agents SDK (Python) | `grantex-strands`            | `pip install grantex-strands`                             | Python                             |
| Anthropic SDK               | `@grantex/anthropic`         | `npm install @grantex/anthropic`                          | TypeScript                         |
| Go SDK                      | `grantex-go@v0.1.10`         | `go get github.com/mishrasanjeev/grantex-go@v0.1.10`      | Go 1.26.1+; workarounds documented |
| A2A Bridge (TS)             | `@grantex/a2a`               | `npm install @grantex/a2a`                                | TypeScript                         |
| A2A Bridge (Py)             | `grantex-a2a`                | `pip install grantex-a2a`                                 | Python                             |
| CLI                         | `@grantex/cli`               | `npm install -g @grantex/cli`                             | TypeScript                         |
| Conformance Suite           | `@grantex/conformance`       | `npm install -g @grantex/conformance`                     | TypeScript                         |
| Event Destinations          | `@grantex/destinations`      | `npm install @grantex/destinations`                       | TypeScript                         |
| MPP Agent Identity          | `@grantex/mpp`               | `npm install @grantex/mpp`                                | TypeScript                         |
| Terraform Provider          | `terraform-provider-grantex` | `source = "mishrasanjeev/grantex"`                        | Go / HCL                           |

<Warning>
  For production MCP enforcement today, use a primary SDK or direct JWKS-backed
  verification at the service boundary. Treat `@grantex/mcp-auth@2.0.2` as
  single-process evaluation software until its documented limitations are fixed.
</Warning>

## Verifiable Intent & Agentic Commerce

Grantex integrates with emerging standards for agentic commerce through W3C Verifiable Credentials and FIDO2/WebAuthn:

<CardGroup cols={2}>
  <Card title="Verifiable Credentials" icon="certificate" href="/features/verifiable-credentials">
    Issue W3C VCs alongside grant tokens. Agents present portable, tamper-evident proof of authorization to any verifier -- no Grantex account required.
  </Card>

  <Card title="FIDO2 / WebAuthn" icon="fingerprint" href="/features/fido-webauthn">
    Passkey-based human presence verification. Cryptographic proof that a real human authorized the agent, embedded in grants and VCs.
  </Card>

  <Card title="DID Infrastructure" icon="id-card" href="/features/did-infrastructure">
    W3C DID document at did:web:grantex.dev. Any party can verify Grantex-issued credentials using published public keys.
  </Card>

  <Card title="MPP Agent Passport" icon="passport" href="/features/mpp-agent-passport">
    Agent identity for machine payments. Issue W3C VC 2.0 passports with spending limits and category restrictions; verifiers can reuse valid cached keys after JWKS retrieval.
  </Card>
</CardGroup>

These capabilities support the **Mastercard Verifiable Intent** specification and similar agentic commerce frameworks. When an agent needs to prove its authorization to a payment processor, merchant, or third-party API, it presents a Verifiable Credential that the receiving party can verify independently using the Grantex DID document.

<Info>
  Verifiable Credentials are opt-in. Pass `credentialFormat: "vc-jwt"` during token exchange to receive a VC alongside the standard grant token. Existing integrations continue to work unchanged.
</Info>

## Enterprise Identity & Access

<CardGroup cols={2}>
  <Card title="Enterprise SSO" icon="key" href="/guides/enterprise-sso">
    OIDC and SAML 2.0 SSO with multi-IdP routing, JIT provisioning,
    group-to-scope mapping, and enforcement, plus an LDAP direct-bind preview
    that does not yet search directories or retrieve groups.
  </Card>

  <Card title="SCIM 2.0 Provisioning" icon="users" href="/sdks/typescript/scim">
    Sync agents and principals from your identity provider automatically via SCIM 2.0.
  </Card>
</CardGroup>

## How Integrations Work

Every integration follows the same pattern:

1. **Wrap** your tool function with Grantex scope enforcement
2. **Pass** a grant token (RS256 JWT) from the authorization flow
3. **Specify** the required scope for the tool
4. The integration **verifies** the JWT signature and claims against the configured JWKS endpoint, then checks `scp`
5. If the scope is missing, the tool **throws immediately** before your function runs

The standalone verification helpers may contact their configured JWKS endpoint
while resolving keys. Keep that endpoint reachable from every protected runtime;
if keys cannot be obtained, protected execution fails closed.