Pre-action authorization

Check risky agent actions before tools execute.

APort guardrails evaluate shell, file, web, MCP, repository, and framework tool calls before the agent can act.

Install my guardrail View supported frameworks

Buyer

Who this is for

AI engineering teams that need practical guardrails without waiting for a full enterprise procurement cycle.

Outcomes

Start in warn/report-only mode, then explicitly move to blocking enforcement.
Use one guardrail model across Claude Code, Cursor, Goose, Codex CLI, Gemini CLI, OpenClaw, LangChain, and CrewAI.
Keep local developer setup simple while offering paid hosted verification for audit and persistence.

Controls

Shell command authorization
File read/write boundaries
MCP tool authorization
Network and web action checks
Framework hook integration

Proof

Terminal warnings or denials
Hosted signed decisions
Local decision output
Framework-specific audit context

Setup

Set up Enforcement the way your team works.

Start agent-first when the agent has repo context, switch to human review when a security owner wants the path, or use the CLI when the integration is already clear.

Recommended

Let my agent prepare Enforcement.

Tell the agent which framework you use: Claude Code, Cursor, Goose, Codex, Gemini CLI, OpenClaw, LangChain, or CrewAI.

agent instruction

install APort runtime guardrails for my agent framework — https://aport.io/skill

Copy this into the agent that already has access to your repo or runtime environment. Require a diff before enforcement or secrets are enabled.

Implementation

Existing quickstart and framework docs

Framework-specific setup stays in quickstart pages and framework pages so install instructions do not drift.

Open my quickstart

Build the control path in layers.

AI agent identity and authorization