Tracecat

What it is

Tracecat is an open source automation platform for security teams. The company describes it as an "AI-native SOAR platform your team can self-host," where SOAR means security orchestration, automation and response. The reader it fits is a security engineer or SOC team that wants to build agents and workflows around its own detections, alerts and tools, and that wants to own where the thing runs. It is closer to a developer-friendly automation platform than to a drag-and-drop business builder, and the pages do not state where the company is based or when it was founded.

The open source product includes agentic AI alongside workflows, cases, tables, integrations, agent presets, skills and a hosted catalog of MCP servers. The company says you can deploy it in your own VPC, on-prem, or on Tracecat's cloud, and that "your prompts and detections stay yours."

What you build and how

The core building block is the agent. Per the company's comparison page, the open source edition includes the agent builder, preset agents, reusable skills, subagents, versioned prompts, sandboxed MCP and Tracecat MCP. Agents delegate to subagents, apply reusable skills and act across the security stack through MCP servers. They ground each investigation in the team's own cases, incoming detections and searchable tables, plus semantic search in connected tools that offer it. The homepage shows examples of agents at work: an endpoint response agent that quarantines a workstation in Falcon, a phishing triage agent that revoked 3 Okta sessions, a cloud agent that applied an STS deny on a deployer role, and an OAuth review agent that staged a revoke for 14 users. These are product demonstrations, not customer claims.

Workflows give you the deterministic side. The homepage describes limitless control flow: loops, if-conditions, parallel subflows and scripts in Python, Bash or Ansible. The site's FAQ says workflows can run from schedules, webhooks and case triggers. Its pitch is to turn prompts into deterministic workflows: run an agent against real alerts inside cases first, then move the repeatable paths into continuous workflows. Typical targets named on the homepage are Tier 1 and Tier 2 work such as phishing, suspicious OAuth grants, EDR malware alerts and cloud findings.

Tracecat MCP lets a coding assistant do the building. The company says it can build and run workflows, agents, tables and cases, and the homepage invites you to connect your coding agent and turn prompts into workflows. A testimonial from a security engineer at Depop says the Tracecat copilot "has changed my life" because they describe an agentic workflow and it builds it for them. The FAQ also answers why not use Claude Code or ChatGPT directly: the company says those are good for reasoning through a playbook, while Tracecat is where the playbook runs, with agents shared across the team and prompts, tool calls, approvals and decisions logged.

Cases and tables are part of the product, not add-ons. Open source includes cases, comments, attachments and custom fields. Enterprise adds agent mentions in case comments and chat, case triggers, approval gates, tasks, metrics, correlation, linked table rows and Git sync of case configuration. Tables are structured data that agents can read during an investigation.

Integrations come in three forms. The homepage claims 500+ enterprise connectors across SIEM, EDR, MDM, IdP and more. It also lists 50+ pre-built hosted MCP servers that agents and workflows can use without integration code, which the company says it curates and maintains. For custom work, the comparison page says Tracecat pulls your Python functions and template integrations from Git in one click, registered once so they run in every workflow and agent.

Control is built in. Sensitive actions can pause for approval, shown on the homepage as an Approve or Deny card, and the company says every approval is logged. Open source audit logs record every prompt, tool call and decision your agents make. Enterprise adds sandboxed nsjail execution on Kubernetes for isolated agent tool calls and Python actions, and workspace Git sync to GitHub, GitLab or Bitbucket for workflows, preset agents, skills, table schemas and case configuration. According to the page, exports open pull requests and imports target one reviewed commit.

The evidence

Pricing is described in structure rather than amounts. The company says the open source edition is free to self-host and that every plan has unlimited workflows, cases and agents, with Enterprise adding multi-tenant workspaces and production controls without per-workflow pricing. The pages shown list no dollar figures, no free-trial terms and no Cloud plan prices, so a team that wants the hosted version has to ask. The site offers Book a demo and Read the docs paths.

The split between editions matters for planning. Open source covers the agent builder, preset agents, skills, subagents, versioned prompts, sandboxed MCP, Tracecat MCP, workflows, cases, tables, integrations and the MCP catalog. Per the FAQ, Enterprise adds tool approvals with a unified inbox, agent mentions in cases and chat, case triggers and metrics, Git sync, custom roles, service accounts and dedicated support. Enterprise also includes a forward deployed security engineer who helps the team build agents around its tools. The comparison page says support is founder led and that customer feedback goes directly into the roadmap.

On deployment, the comparison page says the full control plane self-hosts with Docker. The homepage lists VPC, on-prem and Tracecat cloud as targets. The pages shown name no open source license, no compliance certification and no hosting region, so check the repository and ask the company about those. The homepage says the examples it publishes are linked to NIST CSF 2.0 controls, a useful reference for teams that map automations to a framework.

Testimonials come from named roles at Depop and Neo Financial. They are quotes the company chose to display.

Who it suits and who it does not

Tracecat suits security engineers who are comfortable with Git, code and containers and want AI agents plus deterministic automation without sending their prompts and detections to a vendor-controlled tool. Teams that need to keep case data and automation logic where they host it have a documented route. Teams that need separate workspaces will find multi-tenant workspaces listed under Enterprise. A small team can start on the free open source edition and move up when it needs approvals inbox, roles or Git sync.

It is a weaker fit for a non-technical operations team looking for a general no-code builder for forms, websites or business apps. Everything is aimed at security work, and running the self-hosted version means you operate it. Several of the governance features, including Git sync, case triggers and tool approvals, are Enterprise-only, so a team that expects them in the free edition will be surprised.