AgentTag integrations — connect any AI agent framework
AgentTag connects directly to any MCP-compatible client or A2A agent runtime. Adding AgentTag to an existing agent takes a single CLI command and does not require modifying your existing MCP servers or custom client configurations.
By acting as a proxy gateway, AgentTag brings cryptographic identity, signed policy mandates, and tamper-evident audit logging to all popular LLM tool-calling systems.
Supported AI agent frameworks and clients
Official Clients & MCP
Claude Desktop
Anthropic's official desktop app. Registers automatically in your config and handles step-up approvals.
Claude.ai (Web)
Enforce policy mandates for Claude Web settings using a secure remote MCP server configuration.
OpenAI Assistants
Add a cryptographic layer to OpenAI Assistants via function-calling proxies. Set spend caps and tool limits.
Agent Frameworks
LangChain
Wrap tool execution with signed mandates. Full support for LangChain Python and JavaScript/TypeScript.
LangGraph
Stateful agent flow governance. Assign specific cryptographic mandates to different nodes in the graph.
CrewAI
Provision individual passports to agents in a Crew. Track separate metrics, spends, and logs per agent role.
Microsoft AutoGen
Govern multi-agent conversations. Enforces policy checks at the tool execution level during chat loops.
Agno (phidata)
Secure tool-based agents built with Agno using cryptographic identity and transaction-based audit logs.
Developer Tools & IDEs
Cursor IDE
Configure AgentTag as an MCP server to manage what tools Cursor's composer can run in your codebase.
Cline VS Code
Limit VS Code tool executions, file system writes, and terminal commands by connecting AgentTag.
Windsurf Editor
Integrate as an MCP tool provider to bring unified governance to Windsurf's agentic workflows.
Continue.dev
Brings mandate-based policy controls to VS Code and JetBrains editors via the Continue MCP client.
Infrastructure & Custom Runtimes
Custom MCP Servers
Acts as a gateway for custom-built MCP servers. Compatible with Python and TypeScript SDKs.
REST API
Universal HTTP endpoints to check mandates, log events, and verify keys in any programming language.
Agent-to-Agent (A2A)
Secure communication between different agents. Require signature validation before handing over tasks.
How to connect AgentTag to any MCP client
Integrating AgentTag into your tool-calling client is designed to take less than five minutes. Here is the standard setup flow:
- 01
Install the AgentTag CLI
Open your terminal and run `npm install -g agenttag`. The CLI handles passport creation, certificate exchange, and server settings configuration.
- 02
Mint a passport and configure the client
Run `agenttag mcp add --client <your-client>` where `<your-client>` is 'claude', 'cursor', 'windsurf', 'cline', 'continue', or 'zed'. The CLI registers the passport on your machine and generates configuration configurations automatically.
- 03
Set up your agent mandate
Configure spend caps, tool restrictions, and human step-up approvals for your agent inside the AgentTag control plane. The client immediately enforces these constraints.
- 04
Run the agent
Start your agent or client. It connects to AgentTag's policy gateway, verifying signatures on every single tool execution request. No further changes to your code or tools are required.
What happens when AgentTag is connected?
Once connected, AgentTag wraps all tool executions in a zero-trust envelope. Before any MCP tool (like database access, Slack messaging, or terminal commands) runs:
Policy evaluation
The policy engine checks the agent's signature and verifies that the requested action matches the allowed tools, parameters, and spend limits in its active mandate.
Real-time action
Authorized actions run with low latency. Actions exceeding the mandate are blocked. Actions needing authorization trigger an approval modal on the human's dashboard.
Cryptographic logging
The action's parameters, execution status, and signatures are chained to the audit ledger using SHA-256. This creates a permanent, tamper-evident log.
If you need to revoke an agent, simply run agenttag passport revoke <did>or click 'Revoke' in the dashboard. The agent is cut off immediately, without requiring you to rotate API keys shared by other clients.
Frequently asked questions about AgentTag integrations
How do I integrate AgentTag with Claude Desktop?
+
To integrate AgentTag with Claude Desktop, simply run `agenttag mcp add --client claude` in your terminal. This command mints a new agent passport, registers a default mandate, and appends the AgentTag server to your `claude_desktop_config.json`. Claude will automatically start checking policies on subsequent tool calls.
Does AgentTag work with LangChain?
+
Yes. You can wrap any LangChain tool definition with AgentTag's SDK client. Before a tool executes, the wrapper signs the payload with the agent's passport and validates the action against the policy engine. If allowed, the tool runs as normal.
How does the CrewAI integration work?
+
In CrewAI, you can define different passports for different crew members. For example, a 'research_agent' and a 'writer_agent' get separate passports and distinct spend limits. When they invoke tools, their actions are logged and scoped independently instead of using a shared, broad credential.
Do I need to modify my existing MCP servers?
+
No. AgentTag is a transparent proxy layer. It sits between the MCP client (like Claude) and the MCP server. It handles authorization and signing without requiring code changes to the underlying tools.
Can I connect AgentTag to OpenAI Assistants?
+
Yes. In function-calling models like OpenAI Assistants, you call the AgentTag authorization API inside your tool execution code. This verifies that the assistant's passport has an active, valid mandate for that specific tool and parameters.
Does AgentTag support IDE agents like Cursor or Cline?
+
Yes. By adding AgentTag as an MCP provider inside Cursor or VS Code, you can restrict write access to specific files, block running dangerous terminal commands, or enforce a spend limit on LLM tools.
What is an A2A integration?
+
Agent-to-Agent (A2A) integration enables one agent to securely call another agent. The calling agent signs the request with its cryptographic passport. The receiving agent validates the signature and mandate before executing the requested task.
Can I integrate AgentTag using a REST API?
+
Yes. AgentTag provides standard REST endpoints for checking mandates, logging audit trail records, and verifying keys. You can integrate it with any language (Python, Node.js, Go, Rust, Java, etc.) and any custom agent pipeline.
Connect your first agent in under five minutes
AgentTag is completely free during the public beta. No credit card required. Get started with Claude Desktop, Cursor, or any custom python/TS agent.