LangChain Integration Demo

A runnable demo showing how to wrap LangChain tools with Faramesh governance.

What It Shows

• Block dangerous commands — e.g. rm -rf / is denied while safe commands like ls are allowed

• Policy evaluation — Every tool call goes through Faramesh before execution

• Approval flow — Commands that require approval pause until you approve in the dashboard

Run the Demo

1. Start Faramesh

2. Activate the LangChain policy

The demo uses profiles/langchain_filesystem_policy.yaml or policies/langchain_filesystem_policy.yaml. Activate it, or use a policy that requires approval for shell commands.

3. Run the demo

cd

cd

cd

Or from the Faramesh examples:

cd faramesh-horizon-code
FARAMESH_API_BASE

cd faramesh-horizon-code
FARAMESH_API_BASE

cd faramesh-horizon-code
FARAMESH_API_BASE

4. Approve when prompted

If an action requires approval, open http://localhost:8000 (or your Horizon URL), go to Approvals, and approve or deny. The demo will continue once approved.

How It Works

The demo uses GovernedTool to wrap a shell tool. When the agent tries to run a command:

1. The call is submitted to Faramesh

2. The policy engine evaluates it (allow, deny, or require approval)

3. If denied → the tool raises PermissionError

4. If pending approval → the tool waits (or exits with instructions)

5. If allowed → the tool executes and returns the result

See Framework Integrations for the full LangChain integration guide.