OpenClaw Plugin -- Intaris

An OpenClaw plugin that evaluates every tool call through Intaris safety pipeline before allowing execution.

Features

• Tool Call Guardrails: Every tool call is evaluated via POST /api/v1/evaluate before execution
• Session Management: Creates/manages Intaris sessions mirroring OpenClaw agent sessions, including sub-agent linking
• Escalation Handling: Escalated tool calls enter a polling loop waiting for human approval via the Intaris UI
• Session Recording: Optional audit trail recording of all messages, tool calls, and results
• MCP Tool Proxy: Fetches and registers MCP tools from Intaris-connected MCP servers
• Reasoning Context: Forwards user prompts and assistant responses to Intaris for informed safety decisions
• Fail-open/Fail-closed: Configurable behavior when Intaris is unreachable

Prerequisites

A running Intaris server accessible via HTTP.

Install

openclaw plugins install @fpytloun/openclaw-intaris

Configure

In your openclaw.json:

{
  plugins: {
    entries: {
      intaris: {
        enabled: true,
        config: {
          url: "http://localhost:8060",
          apiKey: "${INTARIS_API_KEY}"
        }
      }
    }
  }
}

Configuration Options

| Key | Env Var | Default | Description | |-----|---------|---------|-------------| | url | INTARIS_URL | http://localhost:8060 | Intaris server URL | | apiKey | INTARIS_API_KEY | (empty) | API key for authentication | | userId | INTARIS_USER_ID | (empty) | User ID (optional if API key maps to user) | | failOpen | INTARIS_FAIL_OPEN | false | Allow tool calls when Intaris is unreachable | | allowPaths | INTARIS_ALLOW_PATHS | (empty) | Comma-separated parent directories for policy allow_paths | | escalationTimeout | INTARIS_ESCALATION_TIMEOUT | 0 | Max seconds to wait for escalation (0 = no timeout) | | checkpointInterval | INTARIS_CHECKPOINT_INTERVAL | 25 | Evaluate calls between checkpoints (0 = disabled) | | recording | INTARIS_SESSION_RECORDING | false | Enable session recording | | recordToolOutput | INTARIS_RECORD_TOOL_OUTPUT | (follows recording) | Record full tool output in events | | recordingFlushSize | INTARIS_RECORDING_FLUSH_SIZE | 50 | Events per recording batch | | recordingFlushMs | INTARIS_RECORDING_FLUSH_MS | 10000 | Recording flush interval in ms | | mcpTools | INTARIS_MCP_TOOLS | true | Enable MCP tool proxy | | mcpToolsCacheTtlMs | INTARIS_MCP_TOOLS_CACHE_TTL_MS | 900000 | MCP tool list cache TTL in ms (15 min) |

How It Works

1. session_start: Creates an Intaris session via POST /api/v1/intention
2. subagent_spawning: Links child Intaris session to parent before sub-agent starts
3. before_agent_start: Forwards user prompt as reasoning context to Intaris
4. before_tool_call: Evaluates every tool call via POST /api/v1/evaluate
   • approve: tool executes normally
   • deny: returns { block: true, blockReason } (blocks execution)
   • escalate: polls for user decision, blocks until resolved
5. after_tool_call: Records tool results for audit trail
6. llm_output: Captures last assistant text for intention context
7. agent_end: Transitions session to idle, sends checkpoints
8. session_end: Signals session completion to Intaris

License

MIT