Skip to content
Agentmart
Submit a tool

Brevet

Change control for what AI agents learn: humans promote, releases are signed, rules can be recalled

Connector by BrightbeamAI · GitHub account verified by the MCP Registry

IndexedApache-2.0Local (stdio)★ 1
Write a review

Chalk Labs

Need help deploying Brevet?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/brightbeamai-brevet

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

Change control for what AI agents learn: humans promote, releases are signed, rules can be recalled

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list, install config and code. Checked Oct 26, 2026.

Tools
6 read · 7 write · 1 delete
Secrets
None asked for
Install scripts
None found
Runs as
Python package

Tools by what they do 14

Read from the source code; the running server may list different tools.

Read 6 · Look things up without changing anything.

  • brevet_active
  • brevet_dawn_pending
  • brevet_harness
  • brevet_recall
  • brevet_status
  • brevet_verify

Write 7 · Create, change, send or run things.

  • brevet_acknowledge
  • brevet_anchor
  • brevet_chap_ingest
  • brevet_dawn_decide
  • brevet_dream
  • brevet_record
  • brevet_release

Delete 1 · Delete, drop, revoke or wipe things.

  • brevet_rollback

Secrets and settings it asks for

Other settings (environment variables or headers).

  • BREVET_AUTO_CAPTURE
  • BREVET_HOME

Network hosts it talks to 4

  • api.github.com · found in code
  • brightbeam.ai · found in code
  • github.com · found in code
  • raw.githubusercontent.com · found in code

Files it touches

  • /Library/Fonts
  • /usr/share/fonts/liberation
  • /usr/share/fonts/truetype/dejavu
  • /usr/share/fonts/truetype/liberation
  • /usr/share/fonts/truetype/liberation2
  • Your home directory (~)
  • ~/.config/brevet/anchors
  • ~/.config/brevet/approvals.yaml
  • ~/.config/brevet/workspaces.json
  • Any path the agent passes to its tools

Tools (14)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

  • brevet_acknowledge

    Confirm that this session has stopped applying recalled rules. Call after brevet_active lists them under "recalled", with their recall ids. Each confirmation is recorded on the evidence chain as the MCP serving point's acknowledgement of that recall.

  • brevet_active

    Return the governed rules of the latest signed release, checked before they are served: the evidence chain must replay intact and match its anchors, the lock must match the digest the chain recorded for that release, the release signature must verify, and each rule's content and conditions must match what was released. Call at session start and follow every rule returned, within its conditions. To receive only the rules for one task, pass its family, domain, role, environment or risk class, and its text as task (a rule's trigger and exclusions are looked for in it); a rule restricted to a role…

  • brevet_anchor

    Write the evidence chain's current head to the anchors configured for this workspace (outside it), so a chain later cut short, rewritten or replaced is detected. Anchoring also happens after every governing step; call this to anchor recent evidence as well.

  • brevet_chap_ingest

    Import CHAP review verdicts as evidence: overrides (diff, rationale and intent_preserved carried through verbatim), rejections (substituting judgments) and approvals (accepted as they were). CHAP is the capture surface; Brevet remains the learning gate. Importing creates evidence only and grants no authority. ``source`` is an audit JSONL file or folder, a coordinator SQLite ``.db`` file, or a served coordinator URL (these two need ``workspace``). A per-source cursor makes re-runs safe, so it can run daily.

  • brevet_dawn_decide

    Apply one dawn-gate decision: promote, hold, reject or re_elicit. The approver must be human:<who> or mission_group:<name>; promotion to the controlled layer needs a mission group. Run only on the user's explicit instruction, with the identity the user gives. When the workspace requires signed approvals, this returns a request for the user to sign.

  • brevet_dawn_pending

    List the candidates awaiting a dawn decision: id, kind, recurrence and title. Eval cases appear too, because their labels can be wrong.

  • brevet_dream

    Run the dream cycle: group recurring overrides into Evidence-layer candidate capabilities and compile eval cases from them. Only what is new is added, from the overrides the manifest's consent scope allows. Candidates have no authority until a human or mission group promotes them.

  • brevet_harness

    Show the harness the latest release locked (files, agent components and library versions, as digests) and which declared harness files changed since, such as an edited prompt or a new skill. A change that is not released is not under change control: tell the user.

  • brevet_recall

    Recall a capability: withdraw it, flag every release whose lockfile contains it, and append the recall notice to the evidence chain. reason_class is one of incorrect, unsafe, consent_withdrawn, superseded, stale, compliance, duplicate or other (duplicate withdraws one byte-identical copy and leaves the content to the copy that stays; every other reason withdraws the content wherever it appears); severity is low, medium, high or critical; action is quarantine, rollback or re_review. Run only on the user's explicit instruction, with the identity the user gives. When the workspace requires signed…

  • brevet_record

    Record one completed task as evidence: the agent's draft and the expert's final (the version actually used). An empty final, or one identical to the draft, means the draft was accepted as it was. family is the task family, the kind of work (for example weekly_summary); tags are comma-separated, and the first one names the problem, so recurring corrections should reuse it. The participant is the human who made the correction and defaults to the workspace owner named in the manifest. Recording creates evidence only; it grants no authority.

  • brevet_release

    Resolve promoted capabilities into capabilities.lock, sign the manifest (Ed25519, covering the lock's digest) and record the release. Trial and production releases need the conservative gate to pass. Give eval_before and eval_after (eval run references) to bind the release to measured runs, whose deltas are then used; otherwise the deltas given here are recorded as attested by the approver, which production releases accept only where the manifest allows it. Run only on the user's explicit instruction, with the identity the user gives. When the workspace requires signed approvals, this returns …

  • brevet_rollback

    Return the agent to an earlier release as a new signed release: its manifest and harness files come back, and its capabilities minus any recalled or decided against since. Run only on the user's explicit instruction, with the identity the user gives. When the workspace requires signed approvals, this returns a request for the user to sign. Harness files added since that release block the rollback unless remove_added sets them aside in the workspace.

  • brevet_status

    Agent status: version, channel, active capabilities by authority layer, recalled capabilities, the dawn queue, signed approvals, the harness, anchors, open recalls and evidence-chain integrity.

  • brevet_verify

    Replay the hash-linked evidence chain, check it against the heads anchored outside the workspace (a chain cut short, rewritten or replaced fails), and check every approval signature on it.

Signals

Stars
1
Forks
0
Open issues
0
Weekly downloads
43
Last commit
Oct 15, 2026 (15 days ago)
Last release
Oct 7, 2026 (24 days ago)
Last package publish
Oct 7, 2026 (24 days ago)

From GitHub (fetched Oct 24, 2026) and the package registry.

Details

Registry name
io.github.BrightbeamAI/brevet
Version
0.4.0
Auth
No auth
Packages
pypi brevet
License
Apache-2.0
First seen
Oct 24, 2026
Last checked
Oct 25, 2026

More from BrightbeamAI

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get Brevet running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying Brevet?

CallGet help
Cite this page

Brevet MCP server. Agent Mart. https://agentmart.dev/connectors/brightbeamai-brevet (last updated 2026-10-25)