Open Source Security Automation for Teams and AI Agents: Scale Your Security Workforce With Agents You Own

COMPANY OVERVIEW

Tracecat, founded in 2023 and backed by Y Combinator (W24), is an open source security automation platform built for security operations, reliability engineering, and IT teams. The platform is licensed under AGPL v3.0, is SOC 2 Type II certified, and runs sandboxed by default. Teams at Saronic, Wealthsimple, SANS, Depop, Acumen, DEVCOM, CyberUp, and Neo Financial run it in production.

The premise is that security teams should own their automation rather than rent it. Tracecat helps analysts become builders and builders become architects, so a team of three can do the work of thirty. Workflows, cases, and AI agents live in the same system, and every definition is stored as YAML that you can version in Git alongside the rest of your infrastructure.

Deployment is yours to choose: self-host the open source edition with Docker Compose for free, run it in your own VPC or on-premises, or use Tracecat Cloud. The enterprise edition adds reserved compute, autoscaling, version control, and bring-your-own-LLM support. Audit logs stay in your Git repository and your VPC.

CORE FOCUS

Tracecat treats deterministic automation and AI reasoning as complementary rather than competing. Workflows handle the paths you already know how to walk: loops, conditionals, parallel execution, error workflows, and schema-validated inputs and outputs. AI agents handle the paths you cannot enumerate in advance, reasoning over evidence and calling the exact tools you allow them to call.

Both are gated by human-in-the-loop approvals with full audit logging. Sensitive actions require sign-off before they execute, and agent runs are bounded by explicit limits on tool calls and model requests so nothing runs away. Execution is sandboxed, with controlled resource access and network boundaries.

Integration is the part most automation platforms make expensive. Tracecat ships over 100 MCP servers covering CrowdStrike Falcon and Microsoft Defender for endpoint, Wiz and AWS GuardDuty for cloud, Okta and Microsoft Entra for identity, Splunk and Elastic for SIEM, and Slack and Gmail for communications. There is no custom integration code to write, and workflows can be generated straight from a natural language prompt using Claude or another coding assistant.

PRODUCTS & TOOLS

Workflows

Deterministic automation that connects triggers to actions and orchestrates work across cases and agents.

• Triggered by webhooks, schedules, subflow actions, and enterprise case, task, and comment events
• Full control flow: conditionals, loops, parallel execution, and dedicated error workflows that receive failure context
• Stored, exported, and versioned as YAML definitions while remaining fully editable in the visual workflow builder
• Draft and published versions are distinct; triggers always run the latest published version

AI Agents

Autonomous reasoning inside workflows, with tool access you control explicitly.

• Model-agnostic across OpenAI, Anthropic, and others, including custom base URLs for self-hosted inference
• An explicit actions list defines exactly which Tracecat actions an agent may call
• Enterprise tool approvals require human sign-off before selected tools run
• Bounded execution by default (15 tool calls, 45 requests) and structured typed outputs rather than raw reasoning

Case Management

Triage, investigation, and reporting in one record, wired directly into automation.

• Ownership, severity, priority, and status consolidated with evidence and follow-up work
• Custom fields for ticket IDs, affected systems, and triage notes across multiple data types
• Threaded comments where analysts and workflows contribute to the same investigation timeline
• Workflows create cases, attach evidence, add comments, and advance status as new information arrives
• An AI copilot summarizes case activity and answers investigation questions inside the case record

Integrations and Deployment

100+ MCP servers, self-hostable anywhere, open source under AGPL v3.0.

• EDR, cloud, identity, SIEM, and communications coverage without writing custom integration code
• Reusable skills built by your team and composed into runbooks
• Free self-hosted edition via Docker Compose; enterprise adds reserved compute, autoscaling, and bring-your-own-LLM
• SOC 2 Type II certified, sandboxed by default, with audit logs retained in your Git repository and VPC