Availsync
Sign inStart free

Changelog

Product updates.

The current focus is first-pilot readiness for coding-agent repo guardrails.

2026-05-27

Claim heartbeat and lost-claim recovery

  • - SDK and MCP guarded runs now opt into claim-level heartbeat by default, with a 5-minute lost-claim timeout.
  • - Availsync can mark a disappeared agent run as lost, clear active file claims, and show heartbeat_missed/lost events in the work timeline.
  • - The new /v1/work/run/heartbeat endpoint records claim liveness without extending the lease or consuming monthly coordination API-call quota.
  • - The work dashboard distinguishes live heartbeat claims, lease-only claims, and lost claims.

2026-05-27

Work event timeline

  • - Work dashboard now shows active claims, lease expiry, renewal count, and current file working sets in one view.
  • - Added a normalized work event timeline for claim events, file working-set events, and work API activity.
  • - Timeline filters let users inspect all work events, a single claim, or a protected resource without reading server logs.
  • - New read-only API endpoints expose /v1/work/events plus claim and resource timeline aliases for debugging agent coordination.

2026-05-22

Simpler agent setup

  • - Added the @availsync/setup CLI as the recommended setup path for Codex, Cursor, and Windsurf MCP configuration.
  • - Claude Code now has a fast path with the native claude mcp add command, so users can connect Availsync without hand-editing JSON or TOML.
  • - Setup output now fills in non-secret values like API URL and agent id while keeping the one-time API key hidden as a placeholder.
  • - Agent Connect docs now show role-based MCP setups for multiple Codex or Claude agents with separate agent ids and priorities.
  • - SDK examples now default to AVAILSYNC_API_URL, use safer timeout/unavailable handling, and show extending claims before the lease expires.

2026-05-22

File-aware coordination

  • - Agents can now declare a file working set during a guarded run, so Availsync can coordinate at repo, project, and file level.
  • - MCP guarded runs can start with file paths and will clearly stop when requested edit/delete paths are already claimed by another agent.
  • - Working-set claims let an agent keep the wider repo run alive while only holding the specific files it is actively editing.
  • - Work claims are now explicit immediate-start guardrails with stricter start-time validation and a database safety net for one active claim per resource.
  • - Node SDK updated to 0.1.8 with stronger types, default withClaim idempotency, auto-extend hooks, and stricter unavailable-mode handling.
  • - MCP package updated to 0.1.5 with guard_work_run as the primary coding-agent entrypoint and legacy work-slot aliases removed.
  • - Work APIs now make /v1/work/run/start the recommended path, with the raw /v1/work/claim endpoint marked as legacy.
  • - Agent API-key verification now uses a short-lived safe cache, and work-claim status/schema cleanup removed unused blocked claim rows.
  • - Setup docs now use run/start, run/extend, and run/finish for new automation examples.

2026-05-14

Paid plan waitlist

  • - Paid Individual and Team plans are waitlist-only during the private pilot.
  • - Checkout stays behind a billing-upgrades flag so paid plans can be enabled quickly later.
  • - Free workspaces remain available for verification, SDK tests, and observe-mode pilots.

2026-05-13

Node SDK

  • - @availsync/node published for guarded repo automations.
  • - withClaim helper wraps start, extend, finish, skip_run, and observe-mode metadata.
  • - SDK docs and pilot guide added for first self-serve integrations.

2026-05-13

Observe-only pilot mode

  • - Agents can run in enforce or observe mode.
  • - Observe mode records would-have-blocked runs without creating active leases.
  • - Overview value metrics separate real prevented conflicts from shadow-only signals.

2026-05-13

Onboarding verification

  • - Dashboard verification shows Availsync resolving a real conflict between two agents.
  • - Verification marks the workspace as verified and cleans up temporary resources.
  • - Activity and audit records distinguish setup traffic from production traffic.

2026-05-12

Work guardrails

  • - Repo and project work claims protect resources before coding agents act.
  • - Runtime APIs support start, extend, finish, idempotency, and clean skip_run responses.
  • - Agent connection, activity, billing, admin, and health views added around the core guardrail flow.