TracecatHQ/tracecat

Open-source security automation platform for teams and AI agents

Tracecat – AI‑native security automation platform

What it is – Tracecat is an open‑source platform that lets security teams build, run, and manage AI agents that automate incident response, case handling, and routine security workflows. It combines a low‑code workflow builder, a case‑management UI, and a “MCP” (Model‑Code‑Prompt) layer that lets LLMs (Claude, Codex, Copilot, etc.) invoke tools and run code just like a human analyst.

Why it matters – Modern security operations spend a lot of time triaging alerts, pulling data from dozens of tools, and writing repeatable playbooks. Tracecat gives teams a single place to:

  • Define agents with custom prompts, toolsets, and “skills”.
  • Hook those agents into cases so the AI can investigate, enrich, and even write a verdict.
  • Orchestrate deterministic steps with a visual DAG builder that runs on Temporal, guaranteeing resilience and scale.
  • Store structured data in tables that agents can query, and expose the same data to external tools via 100+ pre‑built integrations.

Core components

Component Role
Agents & Skills Prompt‑driven LLM bots equipped with tool calls (e.g., API queries, code execution).
Case Management UI for tracking incidents; agents can auto‑populate timelines, IoCs, and verdicts.
Workflows Deterministic DAGs executed by Temporal; can mix AI steps with classic logic.
Tables Simple relational store (PostgreSQL) for entities, observations, and shared variables.
MCP (Model‑Code‑Prompt) Server HTTP/OAuth endpoint that translates LLM prompts into concrete tool calls; 100+ hosted servers are provided.
Integrations Connectors to common security products (SIEMs, ticketing, cloud APIs, etc.).
Sandbox Untrusted code runs inside nsjail or isolated PID runtimes for safety.

How you would use it

  1. Create an agent – write a system prompt, attach tools (e.g., “query Elastic SIEM”, “run a Python script”), and optionally add custom Python scripts from a Git repo.
  2. Define a workflow – chain steps like “fetch alerts → run enrichment script → let agent draft response → post to Slack”.
  3. Open a case – an alert triggers a case; the agent automatically gathers evidence, updates the case timeline, and suggests remediation.
  4. Deploy – run the whole stack via Docker, AWS Fargate, or Kubernetes; an air‑gapped mode is also supported.

Open‑source vs Enterprise – The core platform (backend, frontend, workflow engine) is released under AGPL‑3.0. Enterprise‑only code lives in packages/tracecat-ee and adds multi‑tenant workspaces, fine‑grained RBAC, human‑in‑the‑loop approvals, version‑control sync, and extra monitoring.

Tech stack

  • Backend: Python, FastAPI, SQLAlchemy, Pydantic, uv
  • Frontend: Next.js, TypeScript, React Query, Shadcn UI
  • Durable execution: Temporal
  • Sandboxing: nsjail
  • Storage: PostgreSQL + S3‑compatible object store

Who should look at it – Security operations teams that want to augment analysts with LLM‑driven bots, DevSecOps engineers building automated response playbooks, or anyone interested in combining generative AI with reliable, auditable workflow execution.

All details are taken directly from the repository’s README.

Related

  • Project
    growthxai/output

    A TypeScript framework for building AI workflows and agents that integrates prompt management, tracing, and evaluation into a single codebase to avoid SaaS fragmentation.

  • Project
    Kritt-ai/open-kritt

    An open-source security research platform that orchestrates AI agents to find, validate, and rank vulnerabilities in code through focused, parallelized workflows.

  • Project
    arcjet/arcjet-js

    A runtime security platform for AI applications that detects prompt injection, manages token budgets, and blocks bots and sensitive data leaks.

  • Project
    jd-opensource/JoySafeter

    An AI-native platform for building and orchestrating security agents to automate complex tasks like vulnerability analysis and penetration testing.

  • Project
    useagenthq/useagent

    useAgent is an open‑source, self‑hosted AI coworker platform that lets LLM agents (Claude Code, Codex, OpenCode, Pi) operate inside isolated Linux sandboxes, browse the web, run commands, edit code, and interact with team tools (Slack, GitHub, Gmail, Notion, etc.). It provides a UI with a live timeline, terminal, and workspace, stores every run in PostgreSQL for replay, and supports reusable “skills” and a shared knowledge base. The project is written for the Bun runtime, uses pgvector for embeddings, and is released under AGPL‑3.0 with an optional commercial license.