☿ Mercury Agent v1.3.0 — Mercury Bots

Release v1.3.0

☿ Mercury Agent v1.3.0 — Mercury Bots

Mercury stops being one agent and becomes a fleet. This release introduces Mercury Bots — persistent, persona-scoped agents that run outside your conversation: each with its own persona, private sandbox workspace, skill library, permission scopes, token budget, and a durable job queue. Onboard a researcher, a writer, a watchtower — chain them into multi-level fleets where a lead bot recruits and manages its own crew. They work 24/7, stream their thinking into their own threads, and escalate only what truly needs you.

What's New

• Persistent bot runtime — Bots are long-lived persona agents, not temp sub-agents: bot.yaml manifest + permissions.yaml, private sandbox at ~/.mercury/bots/<id>/, journal-backed thread history that hydrates on reopen. A wake turns them on; they work through a durable job queue that survives restarts.
• Fleets — leads + crews — Promote any bot to fleet lead; it recruits, organizes, and manages its own crew (imperative self-organization). Multi-level hierarchies supported; crew results bubble up through the lead. Onboarding tiers: solo → lead auto-build → lead manual.
• Durable queue + DLQ — Jobs survive restarts, lease-heartbeat while running, classify to timed out on worker death, and replay non-destructively from the DLQ (/bots replay). Nothing dies silently.
• Web cockpit — The local dashboard gained a Mercury Bots section: live roster states, per-bot threads, needs-you escalation badges, and a fleet step view that mirrors the TUI.
• Sandbox + deliverables — Per-bot isolated workspace; final artifacts go out via bot_deliver; retention janitor caps disk; fleet-shared folder for cross-bot files with implicit grants.
• Bundles — /bots export produces a shareable bundle (manifests + personas + permissions + skills; a lead's bundle carries its whole crew). /bots import recreates bots — imported bots start disabled by design; sandbox, journals, and .env never travel.
• Permission tiers at onboarding — ask / allow per capability, set explicitly. Per-bot shell allow-lists flow from permissions.yaml autoApproveCommands; malformed scopes skip + warn, never crash turns.
• Bot-facing tools — dispatch_bot, bot_schedule, bot_deliver, fleet_status, journal_append + native/per-bot skill libraries with auto-synthesis. Bots can schedule their own future runs.
• mercury bots CLI — doctor (exit 1 when actionable), list, storage.

Fixed

• Fleet duplication on reopen; /bots stays inside the bot thread it opened.
• Bot thinking streams live into its thread; fleet speed fixes (single-provider lease wait, batched roster fetches).
• Bot-thread results no longer sliced at 800 chars; durable retries + durable mailboxes close delivery-contract gaps.
• Windows hardening: SQLite queue teardown EBUSY, heartbeat-test EBUSY, retention EINVAL, separator-agnostic traversal guard; post-close queue ops degrade to no-ops.
• JSON backend lease-expiry race closed; doctor flags routines registered in schedules but missing from bot.yaml (the silently-never-fires failure).
• Telegram member installs no longer surface install_skill.

Upgrade

npm:

npm install -g @cosmicstack/mercury-agent@1.3.0
mercury restart

Standalone binary: re-run the one-line installer from mercuryagent.sh, then mercury restart.

No config migrations. Bots are opt-in: /bots create <id> "Name" "Description" onboards your first bot.

Files Touched (highlights)

• src/bots/* — the entire fleet module (new): store, queue (SQLite+JSON), bot-manager, bot-turn, journal, bundle, doctor, persona-template, fleet-*, skill-synthesis, retention, tools/.
• src/core/agent.ts, src/cli/*, src/channels/cli.ts — /bots chat surface, CLI command group, bot-thread region, thinking streams.
• src/web/* — Mercury Bots cockpit section on the dashboard.
• PRs: #131–#139, #117 (RevShare).

Full Changelog: v1.2.7...v1.3.0