237K+ AI events recorded on our own fleet · last 30 days AI events on our own fleet · 30d
Security & governancePut your AI policies into practice.EngineeringGive developers AI tools with clear rules.LeadershipAdoption, ROI and risk in one view.
Works with Cursor · Claude Code · ChatGPT · GitHub CopilotExplore the platform →
Our approachObserve first, then set rules.Why we built itA note from our founder.InsightureThe team behind Tracelet.
DESIGN PARTNERSShape Tracelet with us.Three months, hands-on, in your cloud. →
SecurityHow Tracelet keeps data in your cloud.Shadow AIHow exposed is your organisation?Real-world incidentsWhen AI actions go wrong.BlogPerspectives on governing AI.FAQAnswers to common questions.

One platform. Every altitude.

Every AI coding action, checked before it runs. Developers, security engineers and executives each see it at their own altitude.

Become a design partner

Developers own spend, score and session quality. Team leads see fleet keep rate and output.

Explore engineering
  • See AI's impactAI use beside real delivery outcomes.
  • Coach teams, not peopleSmall teams are hidden so no one is singled out.
  • Consistent setupsEvery repo gets the same approved tools.

Built for how AI coding actually runs.

  1. 1

    Discover every tool

    Assistants, agents and MCP servers on enrolled machines, in one inventory.

    See coverage
  2. 2

    Decide before it runs

    Each action is allowed, held for a person or blocked, by policy.

    Follow a request
  3. 3

    Keep the record

    Every verdict lands with its reason, owner and context, ready to export.

    See the verdicts

Start with the tools your team already uses.

Covered out of the box, with room for the tools you build yourself.

Runs on
  • macOS
  • Windows
  • Linux
On your devices

Engineering AI12+

  • Visual Studio Code
  • Visual Studio
  • GitHub Copilot
  • Antigravity
  • Kiro
  • Claude Code
  • Claude Cowork
  • Windsurf
  • Cursor
  • Devin
  • Codex
  • ChatGPT
In your browser

Browser AI4+

  • ChatGPT
  • Claude
  • Perplexity
  • Grok
Bring your own

Your own tools

  • Internal MCP servers
  • Skills
  • CLIs and plugins

Register them once. They get the same policy and decision record.

One control planeTracelet
  • Shared inventory
  • Central policy
  • Decision records

From first signal to accountable decision.

Follow an AI request from discovery to a decision your team can review.

The agent’s task

“Summarise this week’s customer feedback.”

Example
Step 1 of 4

Find what’s connected.

An unapproved AI tool shows up on three company laptops.

Inventory findingUnapproved tool found

Tracelet spots it and shows who can reach it.

Tool
Unapproved AI tool
Status
Not on the approved list
Seen on
3 laptops
1 / 4

Seven ways a policy can respond.

Stop an action, wait for a person, or let it continue. Whichever your policy picks, the decision is recorded with its reason.

coding agent — ~/app

> Why is the S3 upload failing?

● Read(~/.aws/credentials)

✕ Blocked by Tracelet · Protect cloud credentials

⎿ I can’t read that file. Checking the bucket policy instead.

Decision recordBlockIllustrative

Credential file read prevented.

Actor
A developer
Action
Read a cloud credentials file
Policy
Protect cloud credentials
Outcome
Denied before it ran

Every MCP tool, through one gateway.

Point AI tools at one address. Each call is checked against policy and the caller’s role, then made with their own access.

From

Any MCP client

  • Claude Code
  • Cursor
  • claude.ai
  • ChatGPT
  • Agents and scripts

IDEs, agents and chat apps use the same address.

Tracelet gatewaymcp.yourcompany.com/v1/mcp
Illustrative callgithub.create_pull_requestBruce · Claude Code
Policy
Pull requests allowed
Role
Platform engineer
Access
Bruce’s own GitHub sign-in
Recorded with its decision
To

Your connected tools

  • GitHub
  • Jira
  • Google Workspace
  • Internal APIs
  • Local servers, sandboxed

New servers start at high risk until reviewed.

Works with the systems you already run.

Identity and code context link each record to a person, device and change. Findings go to the people and systems that act on them.

Identity & access

People, teams and roles sync from your directory.

  • Google Workspace
  • Microsoft Entra ID
  • Active Directory / LDAP
  • Other directories

Code context

Each finding links to the change behind it.

  • GitHub
Illustrative findingProduction change
Owner
Bruce · Platform team
Change
payments-api #482
Review required

Service desk

Findings open as review tasks with an owner.

  • ServiceNow
  • Jira
  • Zendesk
  • Freshdesk

Security operations

Findings reach your security tools with their context.

  • Splunk
  • Elastic
  • Datadog
  • Custom integrations
How deployment and privacy work

Give AI room to work. Keep your business in control.

Tell us about your team, the tools you use, and the work you want to protect.

Let’s talk