PALISEC Agent
The native security operator in the console: investigation, policy engineering and operations assistance over your real records.
Use it
Open Console → PALISEC Agent. Ask in plain language, or run a quick investigation (security state, open findings, integration status, explain an operation or error code) that calls a tool directly without a model.
Autonomy levels
- 0 Observe: read tools only.
- 1 Assist (default): also draft, validate, simulate and prepare.
- 2 Confirmed actions: also request an emergency stop, an agent pause, or saving a policy draft. Each is confirmed by a person.
Admins set the level and can disable individual tools or the whole agent. Members see only the tools their role allows.
Evidence and citations
Every tool call is recorded with its input, a minimized output, the records it returned and its duration. Answers cite records as [operation:id]; the console links each one, and flags any citation that no tool returned in that run.
Model provider
Any OpenAI-compatible chat-completions API with tool calling. The deployment uses Groq's free tier (openai/gpt-oss-120b). Configuration is server-side only: PALISEC_AGENT_API_KEY, optional PALISEC_AGENT_BASE_URL and PALISEC_AGENT_MODEL. Without a key, conversation is off and direct tools still work.
API
GET /api/v1/agent/tools,GET|PATCH /api/v1/agent/configurationPOST /api/v1/agent/runs,GET /api/v1/agent/runs/{id},POST …/messages,POST …/continue,POST …/cancel,GET …/eventsPOST /api/v1/agent/tools/{name}(direct tool call)POST /api/v1/agent/actions/{id}/confirm,…/reject,GET /api/v1/agent/activity
Runs advance a few steps per request within a time budget; continue a run until can_continue is false. All routes need a console session.