Changelog

What's new in AgentLedger.

v0.5.0March 31, 2026

The Control Plane Release

  • Policy engine — 6 rule types: rate limits, service allow/blocklists, cost caps, payload regex blocks, human approval requirements
  • Human-in-the-loop approvals — agents pause and wait for dashboard approval before high-risk actions
  • Statistical anomaly detection — hourly baseline computation from 7 days of data, alerts on 2+ stddev deviations
  • Agent evaluations — score actions 0-100 with labels and feedback, trend analytics
  • Rollback hooks — compensating action webhooks fired on agent kill or budget exceeded
  • Live streaming via Server-Sent Events — sub-second action feed with auto-reconnect
  • Trace timeline — waterfall visualization with parallel group detection and I/O inspection
  • Multi-environment support — separate dev, staging, and production data
  • Server-side search and filtering with cursor-based pagination
  • Python SDK — sync and async clients with LangChain and OpenAI integrations
  • Team management — invites, role-based access (owner/admin/member/viewer), audit trail
  • SDK: evaluate(), stream(), waitForApproval(), environment config
  • 10 new docs sections, 42 backend tests, 8 dashboard tabs added
  • Published to npm as agentledger@0.5.0
v0.4.3March 24, 2026

Launch Polish

  • Mobile hamburger menu for landing page navigation
  • Trust & Security section: Open Source, Self-Hostable, Fail-Open, Zero Dependencies
  • FAQ section with accordion
  • Dashboard empty states with actionable CTAs and code snippets
  • Docs: mobile sidebar navigation and scroll-aware highlighting
  • Onboarding step 3 enhanced with quick-start code and next-steps checklist
  • Fixed docs code block rendering (SQL, bash, TypeScript syntax highlighting)
  • Bug fixes: budgets API, rate limiter, revoke keys button
v0.4.1February 18, 2026

Launch

  • AgentLedger is live — dashboard, SDK, and API
  • Core SDK with LangChain, OpenAI, MCP, and Express integrations
  • Real-time dashboard with action feed, cost tracking, and agent controls
  • Budget enforcement with automatic agent pausing
  • Anomaly detection and alerting
  • Webhook support with HMAC-SHA256 verification
  • Kill switch — pause or kill agents instantly
  • Pre-flight checks — block actions before they happen
  • Security hardening: CSP, HSTS, input validation, rate limiting
  • Comprehensive API with 20+ endpoints
  • Self-hosting support with Supabase