Antler

In development

Early access

Changelog

What was built, week by week.

This is a build log. It lists what was built each week, not what anyone can use yet. Antler is in development and not open yet, so there is nothing to install today.

Three more agents recorded before they act, on Linux

On Linux, Codex CLI, Gemini CLI and GitHub Copilot CLI are now recorded before each action, and rules can ask or block. Gemini CLI and GitHub Copilot CLI are recorded the same way on macOS. So is Codex CLI, on macOS and on Windows. On Windows nothing changed for Gemini CLI and GitHub Copilot CLI. The check of the record now also comes as a program of its own, with a written specification of the formats. The recorder got faster, and a workspace can now choose to have an agent recorded after the fact, with no delay and no blocking. Evidence can now leave Antler: outside timestamps, delivery to storage the company owns, and legal holds. This work was merged on 16 and 17 November.

Antler's dashboard with example data: the coverage page of the example company Acme. 76 of 77 checked sessions were fully recorded, and 9 are unknown and not in the percentage. Eight computers are connected, and 99.1% of records are tied to a named person.
The dashboard, with example data: the coverage page, with its count of sessions.
Improvements (35)

Recording

  • On Linux, Antler now records Codex CLI, Gemini CLI and GitHub Copilot CLI before each action, through their hooks. Rules can ask or block there.
  • On macOS, Gemini CLI, GitHub Copilot CLI and Codex CLI are now recorded before each action too, after real runs of the same versions there. On Windows, Codex CLI is as well.
  • Each was checked in real runs of one version, against a local stand-in for the model: Codex CLI 0.169.0, Gemini CLI 0.68.0 and GitHub Copilot CLI 1.0.112.
  • Records of Codex CLI and Gemini CLI sessions now name the input that led to an action.
  • The recorder is faster: about 4 ms for an action at the median, measured on Linux.
  • A workspace can have Claude Code, Codex CLI or Gemini CLI recorded after the fact instead. Nothing then runs before that agent's actions, so there is no delay, and rules cannot ask or block for it.
  • The AI inventory now lists what each AI tool on a computer can reach: its tool servers (MCP), its plugins, and the types of credential it can read. Antler reads the names and types from the tool's settings, never a value.
  • Records of a session start, a prompt and an action about to run now say which program asked.

Rules and alerts

  • Antler checks that the hooks of these agents are still in place on each device, and raises an alert when they are not.
  • Browser controls: for each AI site and kind of account, a workspace can allow, allow and alert, or warn the person and point them to the company account.
  • Blocking a site in the browser is built, but it is offered for no site yet.
  • Alerts can go to a syslog receiver, over TLS, as JSON or CEF. Tested against Antler's own listener only.
  • Rules can name the program, the operation and the resource of common commands, such as a forced push in git or a delete in a cloud command line. A command Antler does not understand matches no such rule.
  • Alerts and approval requests can go to PagerDuty, Opsgenie and Google Chat. These follow each service's published API and were tested against local stand-ins, not live accounts.
  • A rule that asks can send its request to a group of approvers at one destination.

Coverage

  • Each agent Antler knows now has a level on each operating system: enforced, recorded, detected or blind. The coverage page shows it.
  • The signed coverage report has a new format. It states those levels, counts the records at each level, and names the agents found on devices that were not recorded.
  • It adds two ratios: the share of records tied to a named person, and the share of the people and devices on the company's own lists that reported. Without those lists it says not available, never 100%.
  • A report signed in the older format still verifies.
  • The signed coverage report has a third format. For Claude Code sessions on a company account it compares the vendor's own record with the computers' records, session by session, and says which were recorded fully, partly or not at all.
  • It also counts the actions an agent's hooks missed, per agent and version, from the sessions in the report.

Proof and documentation

  • antler-verify is the check as a program of its own, so an auditor can check evidence offline without the recorder.
  • The formats of the records and the evidence are written down in a specification, with test files a second verifier can be held to.
  • An API reference, written from the product's own list of routes.
  • The README and the docs were rewritten, with the performance results of 16 November.
  • A checker now runs on each change to the code. It refuses a sentence in the README, the docs, this site or the dashboard that says more than the product does.
  • A new logo, a pair of antlers, in the dashboard and the browser extension.
  • The format specification is published on this site, with its test files, and a page says what a passing check shows and what it does not.
  • Pages for IT and security reviewers: what the recorder does on a laptop and what it costs, a filled-in security questionnaire, and a plan for a pilot.
  • A workspace can have its log timestamped by an outside timestamp authority it chooses (RFC 3161). antler-verify checks the tokens offline. Tested against a stand-in authority.
  • A workspace can have its records delivered to storage the company owns, as sealed files that antler-verify checks from the files alone.
  • Legal holds: an admin can hold the records of a person, a computer or a time range, with a name and a reason. Held records are skipped by retention and deletion, and each hold is itself recorded.
  • A computer can keep a second key in its security chip (a TPM on Linux, the platform provider on Windows) that signs for its heartbeats, so a copy of its files on another computer shows. This is off by default.
  • A page says how Antler behaves beside another tool's hooks in the same agent, from real runs on Linux with a stand-in second tool.
  • A demo workspace of a made-up company, built by the real recorder, for showing the product. The pictures on this site come from it.
Fixes (3)
  • Turning secrets into tags was slow on long text that repeats one word such as “secret”: a megabyte took 4 seconds. It now takes a small fraction of one.
  • On Windows, a damaged device file is now set aside cleanly, and the computer can connect again.
  • The upload step has a memory limit, so it stays within its budget on a busy machine.
Security (3)
  • On macOS and Windows the recorder now ignores environment settings that could make it trust a certificate other than the operating system's.
  • On Windows, the recorder now looks at what a user's folder holds before it takes the folder over.
  • A choice to record an agent after the fact no longer carries over when a computer moves to another workspace.
Limits (9)
  • On Windows, Gemini CLI is still recorded after the fact, as a log only, and GitHub Copilot CLI is detected only: in a real run of each there, a required check did not pass.
  • Cursor, Windsurf, and GitHub Copilot in an editor or as a desktop app are still detected only.
  • Codex CLI and Gemini CLI let the action run when a hook crashes or times out. GitHub Copilot CLI lets it run when a hook times out.
  • A person, or an agent, can switch Gemini CLI's hooks off. Antler reports that the hooks are disabled.
  • Codex CLI has no confirmation of its own. A rule that asks goes to an approver, or the action is denied.
  • GitHub Copilot CLI keeps no session files that Antler reads, so a call its hooks missed is not counted. Its prompts cannot be blocked.
  • The key in a security chip has run against stand-ins, not yet a physical chip, and it is not offered on macOS. That the key is in a chip is the computer's own statement: Antler does not check a certificate from the chip's maker.
  • The comparison with a vendor's record covers Claude Code on a company account. Sessions that reach the model another way, such as through an API key, and sessions of the other agents are listed as not comparable and are never counted as matched.
  • No real timestamp authority or company storage bucket has been used yet. Nothing here makes a legal claim about holds or timestamps.

Credential recall, company sign-in and the first installers

Records now name the credentials an action touched, so a leaked key can be traced and recalled. Most of the week went into what a company needs: single sign-on, evidence for auditors and packages for device management.

Antler's dashboard with example data: the recall of a GitHub token at the example company Acme. The recall is active, seven of eight computers acknowledged it, and the computers still unproven are named.
The dashboard, with example data: a credential's recall and the computers it targets.
Improvements (22)

Recording

  • Each record now names the credentials its action touched, by type and by tag. The value is never kept.
  • Records written after the fact, from transcripts and session files, carry the same tags.
  • Checked against Claude Code 2.1.322 and 2.1.323. A session from a version that was not checked is still recorded, and is marked as unchecked.
  • If the recorder itself fails, the action runs by default, and the fault is recorded. A workspace can choose to stop the action instead.
  • A browser extension for Chrome and Edge counts use of AI chat sites: the site, how many messages and files, and the kind of account. It keeps no prompt text unless the workspace turns that on. It cannot see Microsoft Copilot in Edge, and it is not in any store.
  • An inventory lists the AI tools found on each device, with the version and the kind of account.

Rules and recall

  • A recall can now stop a credential, an agent or a tool on each device in its scope. It ends in one signed report that names each device where the stop could not be proven.
  • The report lists what it cannot see, such as a program that reads the credential on its own.
  • Approvals can be decided in Slack. Only current members can decide, and that is checked at every click.
  • Alerts can go to Splunk and Microsoft Sentinel, and the dashboard shows whether each one was delivered.
  • A workspace can load an inventory of its AI keys and ask for an alert when a key outside it shows up.

Proof

  • Auditors can download an evidence bundle and check it offline with antler verify, against key fingerprints pinned in advance.
  • An export of search results is signed.
  • Deleting a workspace leaves a signed receipt.
  • The home page checks its example records in your browser.

For companies

  • Members, invitations, API tokens and four roles: admin, analyst, approver and auditor.
  • Each signed-in person can see the records Antler holds about them, and what was withheld.
  • A workspace chooses how long records are kept. Shortening that takes a second sign-in step and a 7-day wait.
  • Backup and restore, with a restore drill, and a warm standby in a second zone.

Install and updates

  • Install scripts for macOS, Linux and Windows, a Homebrew formula and an npm launcher. Each checks the download's SHA-256 before it runs anything.
  • Packages for device management: a macOS installer package, a Windows MSI, and deb and rpm packages for Linux.
  • antler upgrade checks a release's signature before it switches, and can roll back. Automatic upgrades are opt-in.
Security (4)
  • Every read of a record's content, every opened session and every API token request goes into an access trail, sealed into the workspace's chain each hour.
  • Single sign-on through a company's own identity provider: OIDC, Microsoft Entra ID or Google Workspace, tied to email domains the company has proven it owns.
  • SCIM 2.0 adds and removes people from the provider. Removing a person ends their sessions and API tokens at once.
  • On a managed computer, each user enrolls through a helper that runs as root, so no user can read the fleet's enrollment token.
Limits (1)
  • None of these is published. No release has been made, and nothing is in a package registry or a store.

The recorder, the signed chain and the first workspaces

The first week of the build, from 6 November. It holds the base of the product: the recorder for Claude Code, signed records, rules, the server and its dashboard, then sign-in and workspaces.

Antler's dashboard with example data: the live log of the example company Acme, newest first. The newest action, a force-push by Claude Code, needs approval; the ones below it are logged.
The dashboard, with example data: the live log of recorded actions.
Improvements (15)

Recording

  • Antler writes a record of each Claude Code action before the action runs: who, what, and the input that led to it.
  • Secrets in the formats Antler knows become tags on the computer, before a record is written or sent.
  • Codex CLI and Gemini CLI are recorded after the fact, from their session files. That is a log only: nothing can be blocked there.
  • A coverage meter compares the records with Claude Code's own transcripts and counts what was missed.
  • Built for macOS, Linux and Windows, and checked against Claude Code 2.1.318 and 2.1.319.

Rules and approvals

  • A rule logs, flags, asks or blocks. It runs in watch mode, which only records what it would have done, or in enforce mode.
  • The built-in rules all start in watch mode. Nothing is blocked unless a workspace chooses it.
  • An approval goes to named approvers and covers one exact action, not the ones after it.
  • A rule can match on the credential an action touches, and on how: sent to a host, in a prompt, or in the output.

Proof

  • Every record is signed and linked to the one before it, in one chain for each device. The server checks the signature and the chain before it accepts a record.
  • A connected computer keeps its own signed copy of what its workspace received, 30 days by default.
  • antler verify checks that copy offline, without the server.

Workspaces and sign-in

  • antler login shows a code, and you approve the computer in your browser. antler logout sends what was recorded first.

Install

  • antler install sets up the hooks in Claude Code and a timer that uploads. It keeps a backup of your settings and leaves your own hooks in place.
  • antler uninstall removes what the install added.
Security (3)
  • Each workspace has its own database and its own keys. The keys are sealed with a cloud key service.
  • Sign-in with GitHub, Google or Microsoft. There are no passwords.
  • Passkeys, and a second sign-in step before a sensitive change.