ai.shipeasy/mcp

Report whether ~/.config/shipeasy/config.json holds a valid CLI token. Returns { authenticated, project_id, base_url, user_email }.

Launch the PKCE device-auth flow via `shipeasy login`. Opens a browser; blocks up to 5 minutes. Caller should render a 'waiting for browser…' spinner.

Delete ~/.config/shipeasy/config.json — the ONE session shared by the `shipeasy` CLI and every MCP client on this machine. No network call. **This is not restorable from MCP**: re-authenticating needs a browser sign-in in a terminal (`shipeasy login`), which this transport cannot perform, so calling this strands the current task and every other tool until a human signs in again. Do NOT call it to 'reset' or troubleshoot a failing call — a 401/403 is fixed by signing in, not by deleting the credential first. Only call it when the user explicitly asked to sign out, and pass confirm: true to acknowledge that.

Fetch one SDK doc page or snippet. Fetch one feature page (`flags`, `experiments`, …) or nested snippet (`release/experiments`, …), substituting declared {{placeholders}} from caller args.

List an SDK's documentation tree. Fetch an SDK's `/docs/manifest.json` and return the doc tree — feature pages, nested snippet groups, the optional setup-topic map, and whether an installable skill exists.

Fetch an SDK's installable LLM skill. Fetch the SDK's `skill/SKILL.md` (frontmatter intact) so an agent can install it verbatim. The CLI `--install` writes it locally (a consumer fs side-effect).

Get a tracked error. Returns a single tracked error by its id, including the latest occurrence's stack, extras, and consequence, plus `occurrences` — the sampled per-instance detail rows behind the issue (newest first; exhaustive while the issue is small, thinned at volume, capped at 100). Returns…

List tracked errors. Returns a single page of tracked production errors as a **bare JSON array** (no pagination envelope), ordered by `lastSeenAt desc`. Filter with `status`, free-text-search with `q`, and cap the page with `limit`.

Resolve a tracked error. Marks one tracked error `resolved` — the single-purpose "close out" action. Takes no body; it is `PATCH /api/admin/errors/{id}` pinned to `{ "status": "resolved" }`, exposed so tooling can close an error without being handed the full open/resolved/ignored status machine. A res…

Get an error's occurrence series. Returns a bucketed occurrence timeseries for one tracked error (by its fingerprint), read from the `shipeasy_errors` Analytics Engine dataset (near-real-time; ingest lag is seconds). The window bounds are epoch **seconds**; `to` must be strictly greater than `from`. The respon…

Create a translation draft. Stage a new translation draft against a target profile, optionally seeding its keys from a source profile.

List translation drafts. List staged translation drafts awaiting review/publish.

Update a translation draft. Transition a draft's lifecycle state (`open` / `merged` / `abandoned`).

List i18n keys. List keys for a profile, optionally filtered to a name `prefix`.

Push i18n keys (insert-only, or `force` to overwrite). Add NEW keys to a profile. Insert-only by default — existing keys are left untouched and reported back as `skipped`.

Set a key's value and publish it live. Upsert a single key's value into a profile and immediately publish the whole profile (KV rebuild + CDN purge) so the new value is live in one call. The key is inserted when new and overwritten when it already exists. `profile` is a profile name — omit it to target the project'…

Update one i18n key. Overwrite a single existing key's value — the only overwrite path.

Create an i18n profile. Create a locale profile. `name` is the stable handle (e.g. `fr:prod`).

List i18n profiles. Returns every locale profile in the project (e.g. `en:prod`, `fr:prod`).

Publish a profile live. Publish a profile to the CDN — rebuild its KV snapshot + purge the edge. Publishing is PROFILE-WIDE: the whole profile is snapshotted into one KV blob, so the body takes no options.

Archive a metric. Soft-deletes (archives) the metric. Returns `409` if it is attached to a running experiment — stop those experiments first.

Create a metric. Creates an event-backed metric. Pass the query as the DSL string (`query`) **or** the typed IR (`query_ir`) — exactly one. `event_name` must equal the event the query references.

Approve a pending event. Promotes a pending (auto-discovered) event to usable so metrics can query it (`pending` → `0`).

Archive an event. Soft-deletes (archives) the event. Returns `409` if any metric still references it — delete those metrics first.

Register an event. Registers a new event name and (optionally) its typed properties. Only `name` is required.

Get an event. Returns one event's full detail. Resolves by exact id, unique id-prefix, or exact (unique) name.

List events. Returns every catalogued event in the project, including pending auto-discovered names. Each row carries its own `pending` flag, so the unapproved queue can be filtered client-side.

Update an event. Partial update of an event's folder, description, or properties. `name` is immutable.

List experiments using a metric. Returns every experiment that attaches this metric — as `goal`, `guardrail`, or `secondary` — ordered with running experiments first, then by role weight (goal > guardrail > secondary).

Print the metric query DSL grammar. Print the full metric query DSL reference — grammar, aggregation semantics, filter/group-by/ratio rules, and glossed examples — used to author `metrics create --query`.

List metrics. Returns every metric in the project (not paginated) — name, folder, source event, the typed `queryIr`, and the rendered query.

Get a metric's time series. Compiles the metric's typed IR into Analytics Engine SQL and returns the bucketed series over the requested window (near-real-time; ingest lag is seconds). The window bounds are epoch **seconds**; `to` must be strictly greater than `from`. The response echoes the SQL that prod…

Get a metric. Fetch one metric by its id or name, including the rendered DSL query and the typed IR.

Unarchive a metric. Reverses a soft-delete (archive), making the metric live again. Idempotent — unarchiving a metric that is already live succeeds with no effect.

Update a metric. Update a metric's definition — folder, source event, query (`query` DSL or typed `query_ir`), winsorisation, minimum detectable effect, or direction. `name` is immutable. Provide at most one of `query` / `query_ir`.

Ack an item (start a run). Acknowledge a queue item — a person or an AI agent declaring "I'm on this now". Opens a run: stamps who picked the item up and when, assigns them as owner, and moves the item into the matching working status (`investigating_by_ai` for an AI ack, `in_progress` for a human one).…

List connected AI agents. The project's connected AI agents — one per authenticated trigger connector (Claude / Cursor / Copilot / Jules). These are the agent types `ops ack` accepts and the agents a queue item can be assigned to.

Delete an alert rule. Deletes the alert rule. The cron stops evaluating it immediately. Use this (then create a new rule) to repoint alerting at a different metric, since `metricId` is immutable.

List Slack channels. List the project's connected Slack channels — used to resolve an alert rule's notification target.

Create an alert rule. Creates a metric-threshold alert rule. `name`, `metricId`, `comparator`, and `threshold` are required; `windowHours` defaults to `24`, `severity` to `warn`, and `enabled` to `true`.

List alert rules. Returns every alert rule in the project (not paginated). Each rule carries its bound `metricId`, the denormalised `metricName` (or `null` if the metric was removed), the comparator/threshold/window, severity, enabled flag, and delivery target.

Update an alert rule. Partial update of a rule's tunable knobs. `metricId` is immutable — the metric also pins the aggregation, so a body carrying `metricId` is rejected with `409 IMMUTABLE_FIELD`; create a new rule bound to the other metric instead (rule deletion is dashboard-only).

File a bug report.. Files one queue item — a bug report or a feature request — and fires the project's connectors (GitHub issue / Slack). `type` selects which; only the two user-fileable types are accepted (`error`/`alert` tickets are auto-filed by the platform). Returns the new id and per-projec…

Comment on an item. Append a comment to a queue item's thread. The body is markdown (mentions like `@teammate` notify that person; `@shipeasy` asks Jarvis, the AI agent, to reply). Pass `parentId` to reply under an existing top-level comment (one level of threading — a reply to a reply attaches t…

List an item's comments. List the comment thread on a queue item, oldest first. Each comment carries its author (a teammate email, or `system` for a comment authored by Jarvis — the AI agent), its markdown body, and `parentId` for the single level of threaded replies. Removed comments are omitted.

File a queue item (bug or feature request) — pass --type.. Files one queue item — a bug report or a feature request — and fires the project's connectors (GitHub issue / Slack). `type` selects which; only the two user-fileable types are accepted (`error`/`alert` tickets are auto-filed by the platform). Returns the new id and per-projec…

File a feature request.. Files one queue item — a bug report or a feature request — and fires the project's connectors (GitHub issue / Slack). `type` selects which; only the two user-fileable types are accepted (`error`/`alert` tickets are auto-filed by the platform). Returns the new id and per-projec…

List fired alerts. Returns the project's FIRED alerts as a **bare JSON array** (no pagination envelope), ordered by `createdAt desc`. Defaults to the currently-firing ones (`status=active`); pass a `status` to widen to resolved/dismissed history or `all`.

Update a fired alert. Triage writes on one fired alert — the only mutations this surface allows. All body fields are optional (at least one required); only the fields present are changed.

Get one queue item. Fetch a single queue item by its per-project `number` or full id.