Release History

This page tracks public workstation-agent releases published through the signed demo-v0 update catalog.

Canonical live catalog:

https://demo-v0.ai-guard.pro/api/v1/versions

Current public release: 0.40.15.

Summary

Version Status Main change Rollback
0.40.15 current MCP allow audit events, repository scan partial-inspection signals, scan deadlines, clipboard polling hardening, crypto address detection hardening 0.40.14
0.40.14 published configurable AI-agent scan-timeout action: block, warn, allow 0.40.13
0.40.13 published macOS user hooks can read and verify signed policy snapshots 0.40.12
0.40.12 published macOS generated hooks receive the installer-provisioned policy trust store 0.40.11
0.40.11 published bounded AI-agent repository launch scans and recursive hook bypass 0.40.10

0.40.15

Published: 2026-08-24

Rollback: 0.40.14

Changes:

  • MCP proxy emits metadata-only events for allowed MCP requests and responses.
  • Repository scanning reports skipped oversized/unreadable files and partial inspection, so incomplete coverage does not look clean.
  • Inline repository scan paths and manual scan commands have bounded wall-clock deadlines.
  • Clipboard monitor default poll interval is 100ms instead of 500ms; the existing interval override remains supported.
  • Crypto-address detection is hardened with additional corpus and detector cases.
  • Signed Linux, macOS, and Windows artifacts were published.

Verification:

go test ./cmd/station-agent ./apps/controlplane/internal/server ./internal/reposcan ./internal/clipboard ./internal/detectors ./internal/event
node --test apps/admin-ui/test/install-portal.test.js
./deploy/demo-v0/verify-downloads.sh
curl -fsS https://demo-v0.ai-guard.pro/api/v1/versions

0.40.14

Published: 2026-08-21

Rollback: 0.40.13

Changes:

  • Added configurable repository scan timeout action before AI-agent launch.
  • The default remains fail-closed: block.
  • Trusted developer workstations can set TECHCATALYST_GUARD_AI_AGENT_SCAN_TIMEOUT_ACTION=warn or allow.
  • Policy integrity failures, Git state failures, and repository findings remain blocking regardless of timeout action.
  • Non-blocking timeout verdicts are cached so the same repository state and policy identity do not repeatedly burn CPU.
  • Signed Linux, macOS, and Windows artifacts were published.

Verification:

go test ./cmd/station-agent ./internal/shellhook
go test ./...
node --test apps/admin-ui/test/install-portal.test.js
TARGET_OS=linux TARGET_ARCH=amd64 VERSION=0.40.14-smoke ./deploy/macos/smoke-test.sh

0.40.13

Published: 2026-08-20

Rollback: 0.40.12

Changes:

  • Fixed the macOS runtime path where user-level AI-agent hooks could not read the signed policy snapshot.
  • Fetched policy snapshots are now written root-owned and mode 0644.
  • macOS bootstrap placeholder policy snapshots are created mode 0644.
  • Added station-agent ai-agent-gate -trust-store.
  • AI-agent gate resolves trust material from an explicit flag, TCG_POLICY_TRUST_STORE_FILE, or adjacent policy-trust.json.
  • Generated AI-agent hooks pass -trust-store directly.

Verification:

go test ./...

Also verified with a macOS package smoke test, Admin UI install portal test, runtime generated-hook simulation with fake qwen, and live demo-v0 catalog/artifact checks.

0.40.12

Published: 2026-08-20

Rollback: 0.40.11

Changes:

  • Fixed macOS generated full hooks and AI-agent hooks so installer-provisioned policy trust material is passed to all policy-evaluating commands.
  • macOS DMG/PKG packaging renders hooks with -trust-store "${TCG_POLICY_TRUST_STORE_FILE}".
  • macOS package smoke test asserts trust-store wiring.
  • Signed Linux, macOS, and Windows artifacts were published.

Verification:

  • macOS package smoke test.
  • Public install manifest and artifact URL checks.
  • Live demo-v0 catalog update.

0.40.11

Published: 2026-08-20

Rollback: 0.40.10

Changes:

  • Fixed high CPU and launch hangs for qwen, claude, codex, cursor-agent, and cursor on large repositories.
  • Added launch-safe AI-agent scan budgets for file count, file size, file concurrency, Git-history object count, and wall-clock duration.
  • Git-history subprocesses are cancelled when the local scan budget expires.
  • The POSIX AI-agent hook passes bounded scan flags explicitly.
  • Permitted AI-agent processes receive a session-token-bound bypass marker to prevent recursive hook re-entry.
  • Repository scan verdicts are cached across supported AI-agent commands for the same repository state and policy identity.
  • Release tooling publishes Windows zip artifacts alongside Linux/macOS artifacts.

Verification:

  • Focused station-agent and shellhook tests.
  • Cross-compiled macOS arm64 build.
  • Public install manifest and artifact URL checks.
  • Live demo-v0 catalog update.

Operator Commands

Check the live catalog:

curl -fsS https://demo-v0.ai-guard.pro/api/v1/versions

Check update availability from a workstation:

station-agent update-check \
  -control-plane-url https://demo-v0.ai-guard.pro \
  -current-version "$(station-agent version | sed -n 's/.*\"version\": \"\([^\"]*\)\".*/\1/p')" \
  -require-signed

Apply the signed current version manually:

sudo station-agent update-apply \
  -control-plane-url https://demo-v0.ai-guard.pro \
  -current-version "$(station-agent version | sed -n 's/.*\"version\": \"\([^\"]*\)\".*/\1/p')" \
  -install-path /usr/local/bin/station-agent \
  -trust-store "/Library/Application Support/Techcatalyst Guard/policy-trust.json" \
  -require-signed

results matching ""

    No results matching ""