Skip to content
Start free

Automate the routine, and gate the risky steps

Build workflows and AI agents on a typed canvas, with schedule, webhook, chat and alert triggers and actions in the tools your teams use. Put an approval in front of risky steps, which network changes cannot skip, and let Auto-Fix turn production errors into pull requests you review.

Start freeBook a demoRead the workflow docs
  • 62 playbooks
  • 1,152 action packs
  • Approval before change

How it works

From a trigger to a step a person approved

Workflows run on the same engine as Brain and the SOC agents, under the same approvals, tool policy and switch that turns AI off.

  1. 01StartInstall a playbook from the library, describe a workflow to Brain, or begin on a blank canvas.
  2. 02TriggerRun on a schedule, a webhook, a chat message, a new alert, a triage verdict or a new case.
  3. 03BuildWire AI agents, connector actions, branches, loops and sub-workflows through typed ports.
  4. 04GatePut Wait for Approval before steps that change things; network changes cannot run without it.
  5. 05RunCredentials resolve when the workflow runs, within the access of the person who runs it.
  6. 06InspectOpen any run to see each step's input, output and request, kept 30 days by default.

Builder

A typed canvas, with AI where it helps

Triggers, AI agents, connector actions, branches and approvals connect through typed ports, so an agent's model, memory and tools are wired the same way every time. Saved workflows stay drafts until you activate them.

  • AI Agent steps take a chat model and a fallback, memory, tools and an output parser, on Anthropic, OpenAI, Ollama or an OpenAI-compatible endpoint
  • Connector actions from the integration catalog act in the tools your teams use, and higher-risk ones file a proposal for a person instead
  • Failed steps retry up to 10 times or continue with the error, and a separate error workflow can take over

Approvals

Risky steps wait for a person, and the engine checks

Wait for Approval pauses a run until a named approver decides, with the facts they need beside Approve and Reject. A network configuration change, an Auto-Fix publish or an MCP server with write tools will not run without an approval upstream.

  • Approvers by name

    Name who may decide, notify them by webhook, and choose what a timeout means.

  • Proposals, not actions

    Higher-risk connector actions file a proposal for a person instead of acting.

  • Change requests only

    A network change step files a change request with its own approvers.

  • Admins for high risk

    Activating a workflow with high-risk connector actions needs an admin.

Auto-Fix

Production errors become pull requests you review

Auto-Fix traces an error to the code, dependency or deploy behind it, has Brain write and check a fix, and opens a pull request only after an approver says yes. It never merges, and a fix counts only once its merge deploys and the error stays away for an hour.

  • Errors from traces, browsers and logs, plus infrastructure and CI failures when you turn those sources on
  • Path rules and a secret scan on every change, and your own tests when you enable them for a repository
  • GitHub, GitLab and Azure DevOps Services, with a daily budget and limits on attempts per error

Better together

Automation that uses the approvals you already have

Workflows act through the same change requests, response proposals and code reviews as the people on your team.

FAQ

Questions teams ask before they automate

Talk to an engineer
Can a workflow change a system without approval?
Some steps cannot: network configuration changes, Auto-Fix publishing and MCP servers with write tools need an approval upstream before they run, and higher-risk connector actions file a proposal for a person. Other steps, such as a Slack post or an HTTP Request, do what you build them to do, so add a Wait for Approval step where you need one.
Does Auto-Fix merge pull requests?
No. Auto-Fix opens a pull request after an approver says yes and never merges it. If you want merges automated, a separate playbook merges a GitHub pull request after a person approves, and you choose whether to install it.
Which code hosts does Auto-Fix support?
GitHub, GitLab and Azure DevOps Services. Bitbucket is coming soon, and Azure DevOps Server is not supported.
Can other AI assistants run Tracix workflows over MCP?
Not yet. Workflows and agents can call MCP servers you add, but Tracix does not run an MCP server of its own.

Install a playbook and run it once

Pick a playbook from the library, bind your credentials, and use Install and test to watch every step run.

Start freeRead the guide