Grantex
@mishrasanjeev
About Grantex
grantex is the identity, authorization, and audit infrastructure for AI agents — the "OAuth moment" for the agentic internet. We provide a universal SDK and cloud service that lets any AI agent act on behalf of a human with scoped, revocable permissions, cryptographic identity, a
Config
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"grantex": {
"command": "npx",
"args": [
"-y",
"@grantex/mcp"
],
"env": {
"GRANTEX_API_KEY": "your-api-key"
}
}
}
}Tools
17Register a new AI agent with Grantex
List all registered agents
Get details for a specific agent
Update an existing agent
Delete an agent
Start an authorization flow — returns a consent URL for the user to approve
Exchange authorization code for grant token
Verify a grant token and inspect its claims
Revoke a grant token by its JTI
Refresh a grant token using a refresh token (single-use rotation)
List grants with optional filters
Get details for a specific grant
Revoke an active grant
Delegate a grant to a sub-agent with narrowed scopes
Log an audit entry for an agent action
List audit entries with optional filters
Create a principal session for end-user permission management
Overview
What is Grantex?
Grantex is a delegated authorization protocol for AI agents, analogous to OAuth 2.0 for humans. It provides scoped consent, per-agent identity, granular revocation, and spending limits for agent actions. The Grantex MCP server implements OAuth 2.1 + PKCE for MCP servers and serves as an OACP (Open Agent Commerce Protocol) authority for agentic commerce, signing and verifying artifacts while remaining separate from merchant connector runtime or payment execution.
How to use Grantex?
Install the Grantex SDK (npm install @grantex/sdk, pip install grantex, or go get github.com/mishrasanjeev/grantex-go) and configure with an API key. Register an agent, request authorization from a user (via Grantex consent UI), exchange the authorization code for a signed RS256 JWT grant token, then verify the token offline using published JWKS. For MCP servers, use the MCP Auth Server in managed or self-hosted mode with OAuth 2.1 + PKCE.
Key features of Grantex
- Offline token verification using published JWKS (no network call needed)
- Per-agent identity and scoped consent for end users
- Granular revocation of individual agent access
- Spending limits on agent actions (e.g.,
payments:initiate:max_500) - Audit logging for every agent action
- MCP Auth Server with OAuth 2.1 + PKCE support
- Public trust registry and anomaly detection with integrations
Use cases of Grantex
- AI travel agents booking flights and hotels on behalf of users with specific spending limits
- AI email agents that can send emails only after user consent and with scoped permissions
- Deploying AI agents that interact with cloud APIs without exposing full API keys
- Agentic commerce where buyer and seller agents negotiate and execute transactions under verified authority
- Multi-agent systems where one agent delegates to another with controlled scopes
FAQ from Grantex
How is Grantex different from OAuth?
OAuth solves delegated authorization for web apps; Grantex solves the same problem for AI agents, adding per-agent identity, scoped consent, revocation granularity, and spending limits that OAuth does not provide for agents.
What are the runtime dependencies?
The Grantex SDK supports TypeScript, Python, and Go. The MCP Auth Server can be self-hosted. The protocol is fully self-hostable under Apache 2.0 license. No external runtime is required to verify tokens; offline verification uses published JWKS.
Where does Grantex store data?
Grantex operates a cloud service (api.grantex.dev) and a public trust registry at grantex.dev/registry. The protocol is self-hostable, so data can reside on your own infrastructure. Grantex is not a merchant connector runtime and does not store payment or order execution data.
What are the known limits of Grantex?
Grantex is an OACP artifact authority only; it does not execute payments, orders, or POS transactions. It does not act as a toll booth for every buyer and seller message. The MCP Auth Server is separate from the Grantex protocol authority.
What transports and authentication methods are supported?
The Grantex API uses HTTPS with API key authentication for server-side operations. The MCP Auth Server supports OAuth 2.1 + PKCE. Token verification is offline via JWKS. The protocol works over HTTP/HTTPS and can be used with any transport that supports signed JWTs.
Frequently asked questions
How is Grantex different from OAuth?
OAuth solves delegated authorization for web apps; Grantex solves the same problem for AI agents, adding per-agent identity, scoped consent, revocation granularity, and spending limits that OAuth does not provide for agents.
What are the runtime dependencies?
The Grantex SDK supports TypeScript, Python, and Go. The MCP Auth Server can be self-hosted. The protocol is fully self-hostable under Apache 2.0 license. No external runtime is required to verify tokens; offline verification uses published JWKS.
Where does Grantex store data?
Grantex operates a cloud service (api.grantex.dev) and a public trust registry at grantex.dev/registry. The protocol is self-hostable, so data can reside on your own infrastructure. Grantex is not a merchant connector runtime and does not store payment or order execution data.
What are the known limits of Grantex?
Grantex is an OACP artifact authority only; it does not execute payments, orders, or POS transactions. It does not act as a toll booth for every buyer and seller message. The MCP Auth Server is separate from the Grantex protocol authority.
What transports and authentication methods are supported?
The Grantex API uses HTTPS with API key authentication for server-side operations. The MCP Auth Server supports OAuth 2.1 + PKCE. Token verification is offline via JWKS. The protocol works over HTTP/HTTPS and can be used with any transport that supports signed JWTs.
Basic information
Category
Developer Tools
License
NOASSERTION
Transports
stdio
Author
mishrasanjeev
Submitted by
Sanjeev Kumar
More Developer Tools MCP servers

TaskerArmy Agent
TaskerArmyAsk Claude or ChatGPT what Shopify theme optimization tasks are pending on your store, a remote MCP server for TaskerArmy Agent accounts.

directree
coburn2716directree MCP server — query the honest software directory from any AI assistant
Mobbin
MobbinMobbin MCP connects your AI agents to 600,000+ real product screens, so what they build starts with what already works.

Hermoso
HermosoHermoso is a remote MCP server that runs a full marketing operation from your agent: research competitor ads, generate image and video ads, publish to social channels, and build and manage paid campaigns. 718 tools (306

Moxie Docs
Jackalope-DevAutomated codebase documentation for GitHub: searchable docs, MCP context for AI agents, doc drift detection, and Friday Cleanup PRs.
Comments