Daemon outbox now stores resolved target_spec + crypto_box ciphertext
+ nonce per row. Drain worker is a forwarder; no per-row resolution at
drain time. Outbound routing is no longer a placeholder.
Schema additions (additive, NULL allowed for legacy rows): outbox.mesh,
target_spec, nonce, ciphertext, priority. v0.9.0 rows keep draining via
the broadcast fallback so existing in-flight rows finish cleanly.
IPC /v1/send resolves the user-friendly to (display name, hex prefix,
full pubkey, @group, *, #topicId) into a broker-format target_spec at
accept time. DMs encrypt via crypto_box; broadcast/topic/group base64
the plaintext. Hex prefixes (16+ chars) match against connected peers.
CLI thin-client routing extends trySendViaDaemon pattern to peer list
and skill list/get. Three new helpers in services/bridge/daemon-route.ts.
SKILL.md gains ambient mode section: after claudemesh install, raw
claude works for the daemon's attached mesh. Launch stays as the
override path.
Spec at .artifacts/specs/2026-05-04-v2-roadmap-completion.md orders
the remaining v2.0.0 work: multi-mesh daemon (1.26), CLI-to-thin-client
(1.27), mesh-to-workspace rename (1.28), HKDF identity (2.0).
Released as 1.25.0 on npm.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Bump apps/cli/package.json to 1.22.0 (additive feature: claudemesh
daemon long-lived runtime).
- CHANGELOG entry for 1.22.0 covering subcommands, idempotency wiring,
crash recovery, and the deferred Sprint 7 broker hardening.
- Roadmap entry for v0.9.0 daemon foundation right above the v2.0.0
daemon redesign section, so the bridge release is documented as the
shipped step toward the larger architectural shift.
- Move shipped daemon specs (v1..v10 iteration trail + locked v0.9.0
spec + broker-hardening followups) from .artifacts/specs/ to
.artifacts/shipped/ per the project artifact-pipeline convention.
Not in this commit: npm publish and the cli-v1.22.0 GitHub release tag
— both are public-distribution actions and require explicit user
approval.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Pre-launch fix: every visible surface already keyed on slug, so
"name" was a parallel string that only existed to confuse users
on rename ("I renamed but nothing visible changed").
Now slug IS the identifier. claudemesh rename <old> <new> is the
whole rename surface. PATCH /api/cli/meshes/:slug body becomes
{ slug } and the route writes both columns to keep them in sync.
Mesh create derives slug from input.name and stores name = slug.
Pickers drop the (parens). The claudemesh slug verb shipped 30
min ago is removed — merged into rename.
The mesh.name DB column stays for now to avoid touching ~25
reader sites; a follow-up migration drops it.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Slugs are not globally unique (mesh.id is canonical) so the route
only validates the regex and updates the row. CLI refuses a local
collision (two joined meshes sharing a slug would make the picker
ambiguous) and rewrites ~/.claudemesh/config.json on success.
Other peers pick up the new slug on next claudemesh sync.
Server: PATCH /api/cli/meshes/:slug body now accepts { name?, slug? }
— same route, just optional both fields.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The /api/my/meshes/:slug PATCH route was never implemented and
better-auth's enforceAuth middleware can't validate the CLI's
device-code JWT (signed with CLI_SYNC_SECRET, not a better-auth
session). Adds /api/cli/meshes/:slug on the web app — verifies
the HS256 JWT inline, scopes the rename to (slug, ownerUserId).
CLI now calls the new path. Mirrors the cli-sync-token pattern.
Closes the "API error 401: Unauthorized" hit after a successful
claudemesh login.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two new CLI verbs for the file-sharing surface that already existed
on the broker (HTTP /upload + WS get_file/list_files) but was only
reachable through MCP-style docstrings referencing tools that do
not in fact exist:
claudemesh file share <path> [--to peer] [--message "..."]
claudemesh file get <id> [--out path]
Same-host fast path: when --to resolves to a session on the same
hostname, skip MinIO and DM the absolute filepath. The receiver
reads it off disk directly. No bucket roundtrip, no 50 MB cap.
Falls back to encrypted upload when the peer is remote or --upload
is set.
Routes the same-host DM by session pubkey, not displayName, so
sibling sessions of the same member do not trip the v0.5.1
self-DM guard.
Updates the bundled SKILL.md and the MCP server instructions to
reference the real CLI verbs instead of the fictional share_file()
/ get_file() tool calls.
Also: rename.ts now distinguishes mesh-membership from web-account
auth and points users at claudemesh login + the dashboard rather
than emitting a bare "Not signed in".
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Zero-install access to the protocol reference: a fresh `npm i -g
claudemesh-cli` user (or someone running the prebuilt binary) can
now `claudemesh skill | claude --skill-add -` without copying any
files into ~/.claude/skills. The skill markdown is embedded into
the CLI bundle at build time via Bun's text-import attribute.
Also replaces two `<> ALL(...)` raw SQL fragments in the dashboard
unread-count queries with drizzle's notInArray() helper — matches
the same fix already applied to /v1/me/topics in the API package.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
closes the "DM looped back to my own inbox" footgun.
what was happening: peer list returns one row per presence,
including the caller's own session AND its sibling sessions.
the cli filtered out the exact-session row but left siblings
unlabeled — copying their pubkey from peer list silently
targeted your own sibling, and the message arrived in "your
own inbox" because the sender was you.
fix is two-part.
(1) peer list — tag rows whose memberPubkey matches the
caller's stable JoinedMesh.pubkey:
● displayName (this session) — the exact session running
the cli call
● displayName (your other session) — sibling session of
your own member
visually identical otherwise; just the marker.
(2) claudemesh send — refuse a target that exactly matches the
caller's own member pubkey on the mesh, with a hint pointing
at --self for the rare intentional sibling-DM case.
both changes additive — existing scripts that pass display
names or other peers' pubkeys behave identically.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ships v0.5.0 phase 2.
api: three new aggregator endpoints for the per-mesh subsystems
that didn't have one yet.
- GET /v1/me/tasks — open + claimed by default; ?status=all
surfaces completed (30d window). sorted open > claimed > done.
- GET /v1/me/state — every (key, value) row across the user's
meshes, sorted by recency. ?key=foo filters to one key.
- GET /v1/me/memory?q=... — ilike on content + tags, no q
returns the last 30 days. excludes forgotten rows.
cli (1.16.0): task list, state list, recall now route through
the matching aggregator when --mesh is omitted. --mesh foo
still scopes to one mesh (existing behavior preserved).
with this, every per-mesh read verb in the cli either has a
cross-mesh aggregator or doesn't need one. v0.5.0 substrate is
complete.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ships v0.5.0 phase 1.
omitting --mesh on these read verbs now routes through
/v1/me/topics and /v1/me/notifications instead of prompting
the user to pick a mesh. behavior preserved for explicit
--mesh foo.
implementation: resolveMeshForMint helper in commands/me.ts
silently picks the first joined mesh for apikey-mint when
flags.mesh is null. /v1/me/* endpoints resolve the user from
the apikey issuer regardless of which mesh issued the key, so
mint location is irrelevant — only the user identity matters.
help text updated to reflect the new default.
phase 2 (task list, state list, memory recall) needs /v1/me/*
aggregator endpoints first; deferred.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ships v0.4.0 phase 5 — final aggregating verb. v0.4.0 substrate
is complete after this.
api: GET /v1/me/search?q=... matches against topic names +
sender display names + v1 message snippets (base64 decode then
ilike). v2 ciphertext matches only on topic/sender — server has
no topic keys. 30-day window on messages, capped at 50 hits per
category.
cli (1.14.0): claudemesh me search <query> renders topic + msg
sections with inline yellow highlighting. min 2 chars; --json
returns the raw response.
web: /dashboard/search adds an autofocused input + mark
highlighting on every match site (topic name, sender, snippet).
sidebar gets a search entry between activity and invites.
roadmap: phase 5 marked shipped, v0.5.0 default-aggregation
behavior added as the natural next track.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ships v0.4.0 phase 4. final aggregating verb after this is
me search (phase 5).
api: GET /v1/me/activity returns topic messages across every
mesh the user belongs to in a 24h default window (?since=iso
override), excluding messages the caller authored themselves.
"what is happening that i missed", capped at 200.
cli (1.13.0): claudemesh me activity prints a condensed feed
with mesh + topic + sender + relative timestamp + snippet (or
[encrypted] for v2 ciphertext).
web: /dashboard/activity clusters consecutive messages from the
same topic into thread blocks for readability. sidebar gains an
activity entry between notifications and invites.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ships v0.4.0 phase 3.
api: GET /v1/me/notifications aggregates the mesh.notification
table across every joined mesh in a 7-day window (?since=iso
overrides, ?include=all surfaces already-read). returns sender +
topic + mesh context plus a 240-char snippet for v1 plaintext
messages or raw ciphertext for v2 (the dashboard topic-key cache
decrypts client-side).
cli (1.12.0): claudemesh me notifications — terse unread feed
with @ dot, --all to include read, --since for custom window.
web: /dashboard/notifications mirrors the cli view in card form,
adds a notifications entry to the dashboard sidebar between
topics and invites. each card links straight to the topic chat.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
roadmap entry for the me-topics + dashboard-topics ship.
claude.md gets the long-overdue note that apps/web is on coolify
on the ovh vps, not vercel — it does not auto-deploy on push to
gitea-vps the way the broker does, and that mismatch cost a
session of debugging. records the manual deploy command so the
next time we ship a web change we don't rediscover the issue.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
split the v0.4.0 entry into phase 1 (the me/workspace endpoint
+ verb that just shipped in CLI 1.10.0) and phase 2+ (remaining
me topics/notifications/activity/search verbs).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Help text was a wall of monochrome ASCII. Now section headers print
bold-clay, the program title is brand-orange, each verb's syntax is
tinted cyan, and `(alias: ...)` parentheticals are dimmed so they
read as secondary metadata. The styles helper already gates on TTY +
NO_COLOR, so non-interactive output stays unchanged.
Adds .artifacts/specs/2026-05-02-workspace-view.md — the v0.4.0
spec for a per-user virtual workspace that aggregates reads across
all joined meshes while keeping writes mesh-scoped. Roadmap entry
added under v0.3.0.
Previously POST /v1/messages returned the message_queue row id as
`messageId`. Topic posts ARE durable (in topic_message); the queue
entry drains on delivery. Pasting that id into `--reply-to` failed
because the broker validates parents against topic_message, not the
queue. Now `messageId` aliases `historyId` for topic posts; both
`historyId` and `queueId` remain available as explicit fields.
Roadmap and CLI README updated with v0.3.1 reply-to + v0.3.2
multi-session entries.
CLI v1.8.0 on npm. Web stays on v1 plaintext pending the IndexedDB
identity work tracked as phase 3.5.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 1 (notification table) and phase 2 (schema + creator seal)
shipped today. Phase 3 (member-driven re-seal + client-side
encrypt/decrypt) is the cut that actually flips the broker to
ciphertext-only.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the terminal verbs (topic tail / member list / notification
list) explicitly to v1.7.0 so the demo cut summary matches what's
on npm.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Blog post "Agents and humans in the same chat" walks through what
shipped in the v1.7.0 demo cut: topics, REST gateway, real-time
SSE, mentions, notification feed, humans-as-peers. Linked from
the blog index above the original protocol post.
Demo script lays out a five-scene 90-second screen capture: two
terminal agents talking, dashboard topic list, live chat with
@-mention autocomplete, mentions feed cross-platform, close.
Production notes + distribution checklist included.
Marketing screenshots and the actual recording are still TODO.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Updates v1.6.x and v1.7.0 sections with concrete endpoints + client
behaviour for what landed this session. Bridge smoke test and
/v1/peers humans remain open under v1.6.x.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Earlier wording claimed --dangerously-load-development-channels "goes
away" at v3.0.0. That overstated what we know. Some opt-in mechanism
is always required for Claude Code to accept external runtime events
from a third-party process — that's a security invariant, not a quirk
of today's flag.
What changes at v3.0.0 is the FORM of the opt-in (stable settings
entry, native transport subscription, etc.), not its existence. The
"dangerously" / "experimental" / "development" framing is what
disappears, because the underlying API graduates from experimental
to stable. The flag itself, or its successor, lives on as a normal
config entry that claudemesh install writes once.
Public roadmap and internal spec both updated to reflect this.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Public docs/roadmap.md gets the v1.6.0 cut moved to shipped, drops the
v0.2.0-as-next section in favor of a v1.6.x patch line + v1.7.0 demo
cut + v2.0.0 daemon redesign + v3.0.0 native-channels migration target.
Items that were in v0.2.0-next migrate down: gateways and tag routing
land in v0.3.0 alongside per-topic encryption and self-hosted broker.
The detailed strategic version lives at
.artifacts/specs/2026-05-02-roadmap.md — schedule, cost estimates,
migration paths, deliberate exclusions, the load-bearing principle for
the daemon shift ("the user is the unit, not the Claude session").
The public file stays marketing-tone; the artifact captures internal
planning.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The v0.2.0 backend cut. Topics, API keys, REST /api/v1/*, and bridge
peers — all in one CLI release. Adds three new verb namespaces:
topic (channel pub/sub), apikey (REST client auth), bridge (cross-mesh
forwarding).
Also pins @claudemesh/sdk as a workspace devDependency so the bridge
implementation is bundled by Bun at build time and doesn't leak into
the npm tarball's runtime deps.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Backwards compat shim (task 27)
- requireCliAuth() falls back to body.user_id when BROKER_LEGACY_AUTH=1
and no bearer present. Sets Deprecation + Warning headers + bumps a
broker_legacy_auth_hits_total metric so operators can watch the
legacy traffic drain to 0 before removing the shim.
- All handlers parse body BEFORE requireCliAuth so the fallback can
read user_id out of it.
HA readiness (task 29)
- .artifacts/specs/2026-04-15-broker-ha-statelessness-audit.md
documents every in-memory symbol and rollout plan (phase 0-4).
- packaging/docker-compose.ha-local.yml spins up 2 broker replicas
behind Traefik sticky sessions for local smoke testing.
- apps/broker/src/audit.ts now wraps writes in a transaction that
takes pg_advisory_xact_lock(meshId) and re-reads the tail hash
inside the txn. Concurrent broker replicas can no longer fork the
audit chain.
Deploy gate (task 30)
- /health stays permissive (200 even on transient DB blips) so
Docker doesn't kill the container on a glitch.
- New /health/ready checks DB + optional EXPECTED_MIGRATION pin,
returns 503 if either fails. External deploy gate can poll this
and refuse to promote a broken deploy.
Metrics dashboard (task 32)
- packaging/grafana/claudemesh-broker.json: ready-to-import Grafana
dashboard covering active conns, queue depth, routed/rejected
rates, grant drops, legacy-auth hits, conn rejects.
Tests (task 28)
- audit-canonical.test.ts (4 tests) pins canonical JSON semantics.
- grants-enforcement.test.ts (6 tests) covers the member-then-
session-pubkey lookup with default/explicit/blocked branches.
Docs (task 34)
- docs/env-vars.md catalogues every env var the broker + CLI read.
Crypto review prep (task 35)
- .artifacts/specs/2026-04-15-crypto-review-packet.md: reviewer
brief, threat model, scope, test coverage list, deliverables.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- info/inbox commands → unified render.ts
- install route: drop in-memory counter, rely on PostHog + structured logs
- docs: roadmap, CLAUDE.md reflect alpha.31 state
- tests workflow now also builds + smoke-tests the CLI bundle
- homebrew tap bootstrap kit in packaging/homebrew-tap-bootstrap/
(README + copy of the formula template for dropping into the tap repo)
- upstream Claude Code issue draft for rich <channel> UI
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add the foundation for deploying and managing MCP servers on the VPS
broker, with per-peer credential vaults and visibility scopes.
Architecture:
- One Docker container per mesh with a Node supervisor
- Each MCP server runs as a child process with its own stdio pipe
- claudemesh launch installs native MCP entries in ~/.claude.json
- Mid-session deploys fall back to svc__* dynamic tools + list_changed
New components:
- DB: mesh.service + mesh.vault_entry tables, mesh.skill extensions
- Broker: 19 wire protocol types, 11 message handlers, service catalog
in hello_ack with scope filtering, service-manager.ts (775 lines)
- CLI: 13 tool definitions, 12 WS client methods, tool call handlers,
startServiceProxy() for native MCP proxy mode
- Launch: catalog fetch, native MCP entry install, stale sweep, cleanup,
MCP_TIMEOUT=30s, MAX_MCP_OUTPUT_TOKENS=50k
Security: path sanitization on service names, column whitelist on
upsertService, returning()-based delete checks, vault E2E encryption.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
NPCs are mesh data (skills, memory, state), not peers. One API call
per interaction, 3 coordinator peers per faction. Game connector
assembles context from mesh and calls any LLM on demand.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The mesh is the communication fabric, not the simulation engine.
SimController pattern: external controller drives tick loop, computes
visibility, sends observations to peers, collects actions.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
changelog-20260407.md: full implementation details for 21 features
vision-20260407.md: slimmed to shipped summary + remaining items
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
17 of 22 items done, 2 partial. Updated all section headers and
added implementation notes with commits and timestamps.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Pairs with claudemesh-2's new /join/[token] landing page. Users can
now paste a clickable HTTPS URL instead of the dev-only ic:// scheme.
apps/cli/src/invite/parse.ts — new extractInviteToken() handles
four input formats before handing the raw base64url token to the
existing parseInviteLink pipeline:
- https://claudemesh.com/join/<token> (primary, clickable)
- https://claudemesh.com/<locale>/join/<token> (i18n prefix)
- ic://join/<token> (still supported, dev)
- <raw-token> (last resort: bare base64url)
User-facing strings updated to the HTTPS form:
- cli help: "join <url>"
- install success message
- list (no-meshes) hint
- MCP server "no meshes" error
- README.md primary example
- docs/QUICKSTART.md Path A + Path B
Verified extractInviteToken() on all 4 formats — each returns the
same base64url token → same broker /join lookup.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
