v0.8.0

Breaking changes

• web — Replace the workflow memo editor with the conversational-first pipeline shell — a copilot ComposerPane beside a TriggerCard + PipelineStrip pane with inline step inspectors, ghost proposals, per-step tests, a run overlay, and a linkable Runs tab (run history plus a live step timeline with drawers that embed agent-step child-run transcripts).
• control-plane, db, shared, web — Rebuild workflows as control-plane-interpreted pipelines (TRIGGER → STEPS) — a v2-only config with tool/infer/agent steps plus for_each/branch/filter/state, a run_steps execution ledger with replay-based crash recovery and per-workflow durable state, agent steps dispatched as mode:"task" child runs with structured output, explicit onComplete.slackReply delivery, pipeline.* run-stream events, new run/steps/state/step-test routes, granular copilot step tools with connection-tool lookup, and no v1 compatibility (the renderTaskMessage dispatch path is removed and old-shape drafts require a stack reset).

Features

• control-plane, db, shared, web — Make MCP OAuth connections actually work: the health probe now presents the broker's access token through getAccessToken instead of dialling unauthenticated and reporting the server's correct 401 as a rejected credential (so OAuth connections finally populate their tool cache), an oauth row is no longer probed before consent, hasCredentials follows the grant rather than the auth type, discovery is scheme/issuer/resource-validated with challenge-first scopes and offline_access, only invalid_grant retires a grant while a declined or abandoned re-consent leaves a live one untouched, consent results post to the SPA origin (new optional PUBLIC_WEB_URL, defaulting to PUBLIC_APP_URL) carrying a sanitized failure reason each surface explains in the same words, pending flows are bound to the user who started them, and catalog OAuth presets now declare their client-identity strategy — adding an operator-configured pre-registered client mode (MCP_OAUTH_<PREFIX>_CLIENT_ID) and removing the Vercel preset, whose authorization server rejects dynamic registration outright.
• control-plane, web — Harden the MCP OAuth broker after review: provider error strings are narrowed to a closed RFC vocabulary before they can reach last_error or a DTO (an authorization server could otherwise echo back the refresh token, authorization code or client secret it had just received), the consent callback's promotion and failure bookkeeping are guarded by an optimistic-concurrency stamp so a flow that loses its race to a URL or auth-mode change discards its tokens instead of resurrecting a stale grant, MCP_OAUTH_<PREFIX>_ISSUER is now required and exact-matched with no fail-open, a stored dynamic-registration client whose issuer was never recorded is re-registered rather than replayed at a new authorization server, and permanently-dead refreshes (invalid_client, unauthorized_client, invalid_scope, unsupported_grant_type) retire the grant as auth_required instead of being retried forever as transient.
• control-plane, db — Stage an MCP OAuth consent flow's discovered endpoints and client identity on the new connection_oauth.pending_flow column and promote them onto the grant only when a token exchange succeeds, so a re-consent that is declined, abandoned, or rejected can no longer repoint a live grant's token and revocation endpoints at an authorization server the MCP server nominated mid-life.

Fixes & maintenance

• control-plane, db — Confirm a Stop's remote cancel from eve's own stream: runs.turn_id is the only acceptance proof (written when the tail attributes the run's own turn.started, in send order), the live tail sends a turn-qualified cancel or nothing and stays on the stream in observation mode until the run's own turn boundary, a session-terminal answer, or a newer run's turn_id clears remote_cancel_pending_at (never eve's 202 to a pre-turn cancel, never a synthesized running successor or leftover events), the no-tail settlement, the periodic sweep and boot reconciliation re-open observation instead of chasing eve unqualified, and an obligation unconfirmed past REMOTE_CANCEL_OBSERVE_MS (default 10 min) is declared unresolved on the row (remote_cancel_unresolved_at, migration 0017) rather than silently cleared.
• control-plane — Await the reply-delivery ledger's failure settle instead of returning its promise, so a settle write that rejects during shutdown surfaces as a logged delivery failure rather than an unhandled rejection in the control-plane process.
• control-plane — Close five read-then-act ordering windows in dispatch and recovery: a replayed for_each failure verdict is now authoritative over the fresh resolution's shape (a recorded fan_out_exceeded is never reported as items_not_array), every eve dispatch path persists the eve session id BEFORE its post-eve terminal recheck so a Stop racing the create always finds an id to remote-cancel (the fresh-chat and post-reset create routes gain the same recheck), the session-busy predicate counts a canceled run whose dispatch may still be in flight on an eveless session (transient session_busy instead of a second eve session), the post-eve abandon skips its unqualified session-level cancel when a newer run already owns the session, and boot reconciliation's eveless-session close is an atomic guarded UPDATE that leaves a session untouched if it gained its eve id between snapshot and close.
• control-plane — Close three dispatch/recovery crash windows: a crashed for_each now replays its RECORDED resolution verdict (a doomed over-maxItems/non-array loop fails fan_out_exceeded/items_not_array on recovery instead of re-resolving against moved state and executing), the dispatch-attempt marker (runs.started_at) alone is now the recovery authority (boot reconciliation never tails a marker-null run, every chat/agent dispatch arms it pre-send, and a stillborn session:"thread" continuation is safely re-sent into its established session), and a Stop racing an in-flight eve create is handled at the source (the dispatch re-reads the child run after the create returns and remote-cancels with the fresh session id, the Slack thread-claim eviction spares marker-armed possibly-mid-dispatch holders, and boot reconciliation closes abandoned eveless marker-set sessions to free their thread claims).
• control-plane — Move the pipeline runner's per-run driver lock onto a dedicated, bounded lock pool (DB_PIPELINE_LOCK_POOL_SIZE, default 32; exhaustion is the typed transient skip lock_pool_exhausted taken before any run row exists — 503 pipeline_lock_pool_exhausted on webhook/manual ingress — instead of pinning root-pool connections until ten long pipelines wedged the whole control plane), clear a Stop's durable remote-cancel obligation only on a confirmed outcome (eve acknowledged, eve session_not_active, superseded, or nothing to chase — a transport failure now retains remote_cancel_pending_at, and a live-tail cancel that fails in transport records it exactly like a skipped one), make the deferred cancel chase back off across a saturated lock pool for its whole bound, and add a periodic advisory-locked remote-cancel sweep (REMOTE_CANCEL_SWEEP_MS, default 60 s) so a healthy process finishes those obligations without a restart.
• control-plane — Close four run-lifecycle defects: a live-tail Stop whose remote cancel was skipped or failed in transport now writes its durable remote_cancel_pending_at obligation inside the same CAS that finalizes the row canceled (never a second statement after it), "superseded" now requires a successor that provably reached eve (running/waiting, or terminal with observed events — a merely queued successor retains the marker and the chase retries, in both the guarded cancel and the post-eve recheck), a periodic pipeline-recovery sweep (PIPELINE_RECOVERY_SWEEP_MS, default 60 s, replica-elected and lock-gated) re-adopts interrupted pipeline runs left locked at boot instead of leaving them active until a restart, and the manual "Run now" route answers the transient 503 pipeline_lock_pool_exhausted for a pinned lock pool instead of collapsing it into 409 run_overlap_skipped.
• control-plane — Serialize each agent session's dispatch decisions under a per-session Postgres advisory lock (held across the eve call and its settlement, released by a crash with its connection): admission, the canceled-dispatch abandon, and the boot sweep's eveless close no longer race as separate read-then-acts — a canceled dispatch's accepted turn is always remote-canceled instead of leaking when a successor slipped in, the unqualified session-level cancel can never land after a successor's turn started (the cancel routes' remote leg is lock-guarded and deferred, never dropped, while a dispatch is in flight), the sweep skips lock-held candidates and its guarded UPDATE atomically re-asserts the ledger, and an eve create that fails after arming the dispatch-attempt marker (a client timeout racing eve's 202 included) closes the still-eveless session on any terminal outcome so the next message mints a fresh session instead of a second live turn beside an unobservable one.
• control-plane, db — Close six defects in the per-session dispatch lock: the lock now rides a dedicated Postgres pool (DB_LOCK_POOL_SIZE) with a bounded reserve so concurrent dispatches can no longer deadlock the root pool, a fresh session's id is pre-minted so its lock is taken before the claim transaction (a timeout creates nothing) and continuations re-read their session under the lock with CAS-only status writes (no stale-snapshot resurrection of a closed row), a lock-holding dispatch heals an abandoned eveless session inline instead of answering session_busy until a restart, a Stop on a live tail awaits its remote cancel u...

v0.7.0

Features

• control-plane — Make reasoning effort actually reach OpenRouter from the control plane: align the AI SDK pins with the compiler's version matrix (retiring a stale @openrouter/ai-sdk-provider alpha that silently dropped every reasoning route), route the effort per provider through the new model/reasoning.ts, and add COPILOT_REASONING_EFFORT (default off) so the copilot can ask for one too.

Fixes & maintenance

• web — Fix signing in requiring two attempts and workspace content staying blank until a reload, by moving SPA identity off Better Auth's React hooks onto a viewer query gated in the router; the cached query data of a previous account is now dropped whenever the signed-in principal changes, including from another tab, a session revoked elsewhere is noticed on refocus or reconnect, every auth request behind the gate is bounded so a hung server cannot wedge navigation, and accepting a workspace invitation can no longer strand you on a retry that re-reads the invitation it already consumed.

v0.6.0

Breaking changes

• compiler, control-plane, db, design-tokens, shared, site, web — The agent content hash now keys on the agent's stable id instead of its slugified display name, so every agent rebuilds once on its next publish and duplicate agent names are legal; alongside it, publishing no longer blocks the editor and its build watch follows you across the whole app, the editor separates unsaved from unpublished changes, chat sessions enter instantly and title themselves (falling back to the thread's first message, which the session list now carries), tool calls and a context-budget meter replace raw slugs and build identity in the thread, clearing or compacting the context leaves a divider that survives a reload while a compaction with nothing to summarize says so, and the copilot shows its steps and reasoning, can set an agent's name and description and sees both as the editor currently holds them, can apply edits without the accept gate, and reports an edit that failed to apply as failed.

Features

• site — Bring the marketing/docs site back in sync with everything shipped since the agents-first pivot: a new Chat page covering the run timeline, Stop, the message queue and context controls; reasoning effort documented across Models, Agents and Overview; corrected tool-filter scope on Context & MCP; guarded egress and OAuth token custody on Security; the search index on Architecture and Deploy your own.
• site — Flesh out the docs site into a comprehensive reference: expand all seventeen existing pages past their placeholder state, add a Guides section with five task-based walkthroughs and a Reference section with a glossary, limits and defaults, troubleshooting, and keyboard and accessibility notes, and add new Building pages for Skills and Settings.
• site — Prerender every landing and docs route to static HTML with per-page title, description, canonical, OG and JSON-LD metadata, and add a generated sitemap.xml, robots.txt and llms.txt, real 404 statuses, a 301 for /docs, and noindex on preview deploys.

v0.5.0

Breaking changes

• control-plane, db, shared, web — Rebuild the connections domain on new connections tables with a curated connector catalog, Meilisearch-mirrored community registry search, and unified create routes replacing the registry install flow.

Features

• compiler, control-plane, db, shared, web — Add the MCP OAuth 2.1 broker: path-aware discovery, CIMD/DCR client identity, PKCE popup consent, envelope-encrypted tokens with single-flight central refresh, an agent-facing token route with audience-derived version binding, oauth codegen via getToken, and OAuth connectors in the catalog.
• control-plane, web — Add connection health probes, cached tool discovery with a checkbox tool picker, a connection detail surface with test-connection, and a hardened SSRF-guarded egress path for all caller-influenced control-plane fetches.

v0.4.0

Features

• web — Move the chat Stop control onto the composer's send button and queue messages typed during a run, merging them into one send when the session frees.
• web — Render a run as an ordered timeline of work and speech segments so reasoning accumulates instead of overwriting, interim narration renders in chronological position, and streaming text never relocates after it has rendered.

Fixes & maintenance

• worker — Stop the artifact-cache boot scan from killing worker startup when an entry disappears between readdir and stat.
• web — Stub every server-only packages/shared module for the browser, fixing the node:crypto crash that made the SPA fail to load, and add a guard test so a new server-only module in the shared barrel can no longer break the client bundle silently.
• web — Replay a rich-text editor focus request that arrives before the editor instance exists instead of silently dropping it.

v0.3.0

Breaking changes

• compiler, control-plane, shared, web, worker — Upgrade eve 0.19.0 → 0.31.3. Sessions are now ID-addressed (session API v2), continuation tokens are gone, and stop plus context controls are available. Every published agent must be republished to migrate.
• compiler, control-plane, db, shared, web — Reasoning effort is now set per preset and inherited by agents, and the default model set has changed. Existing agents pick up the new defaults on their next publish.

Features

• control-plane — Adopt changesets for releases: merging the chore(release): version packages PR now computes the version, writes CHANGELOG.md, tags vX.Y.Z, cuts the GitHub Release, and builds the GHCR images in one workflow run.
• web — Replace the chat markdown renderer with Streamdown, including a streaming caret and E1-themed code blocks.
• design-tokens, web — Replace the CodeMirror prompt editors with Tiptap, including the E1 token work the new editor surface needs.

Fixes & maintenance

• control-plane — Align the production images' node with packages/compiler/versions.json and add a guard so the two cannot drift.
• control-plane — Pin the host toolchain (bun, node, wrangler) in mise.toml and install it in every CI job via mise-action, so all lanes run the same versions.

v0.2.0 — agents-first

Agents-first re-architecture (breaking)

The platform's center of gravity moves to Agents — a persona, a model, and the context/tools they're equipped with. You chat with Agents directly, and workflows become standing delegations: trigger → agent → instructions, published instantly with no builds.

Highlights

• Agents are the compile unit: each published Agent builds to its own versioned eve artifact (world DB ag_v_<hash>, per-version JWT agent-version:<hash>); artifacts expose only eve's default channel
• Dispatch-time instructions: the control plane renders workflow instructions + trigger data into the task message per run — editing a delegation never rebuilds anything
• Slack replies from the control plane (at-least-once with boot recovery) and working schedule triggers via an advisory-locked cron ticker
• New SPA surfaces: top-level Agents section with the agent editor, agent-picker chat with provenance chips, single-column delegation editor with per-trigger test runs, copilot on both editors
• Compiler 3.0.1, migrations 0005–0006, full acceptance/e2e suites rewritten agents-first
• Messaging pivot across README, docs, and invisiblestring.io: describe / build / equip / delegate

Breaking / upgrade notes

• No data migration. Migration 0005 is destructive by design (pre-production): existing deployments must reset their product database (and world DBs / object store) — docker compose down -v on dev stacks; for prod compose follow docs/DEPLOY.md with a fresh data volume
• All content hashes re-key (compiler major): first publishes after upgrade rebuild from scratch
• API surface renamed throughout (agents lifecycle routes, session/workflow DTOs); see docs/superpowers/specs/2026-07-10-agents-first-redesign.md for the full decision record

Images: ghcr.io/heysanil/invisible-string-{control-plane,worker,web}:v0.2.0 (built by the release workflow; set IMAGE_TAG=v0.2.0 in prod compose).