Docs

Tools reference

Every tool of the Sypnia MCP server: what it is for, its input, and an example call with its result.

Applies to EngineCloud At launch

View as Markdown
On this page
  1. health
  2. problems
  3. problem
  4. problem_prompt
  5. sessions
  6. session
  7. search
  8. performance
  9. uptime
  10. resolve_problem
  11. Resources
  12. Prompts

Engine Cloud The Sypnia MCP server has ten tools. Nine only read. One writes: resolve_problem. The tools are the same on the remote server of Sypnia Cloud, on the one a self-hosted engine serves, and on the local @sypnia/mcp. Their machine-readable contract is spec/mcp.json in the docs repository.

Your client shows the tools under the server’s name, sypnia. Depending on the client that reads as sypnia.problems, sypnia:problems or mcp__sypnia__problems. It is the same tool.

ToolUse it toWrites?
healthAsk «what is failing in production?»No
problemsList problemsNo
problemGet one problem with everything needed to fix itNo
problem_promptGet the ready prompt of a problem, as textNo
sessionsList the sessions a problem affected, or recent sessionsNo
sessionRead one session, event by eventNo
searchFind a problem, session, person, release or page by textNo
performanceSee slow pages, slow or failing API calls, and what got worseNo
uptimeSee whether the app answersNo
resolve_problemMark a problem as resolved after the fixYes

Rules that hold for every tool:

  • One project. The API token belongs to one project, so no tool takes a project.
  • Production by default. Tools that read data take environment; leave it out for production.
  • Results are JSON. Each result comes as structured content and, for clients that only read text, as the same JSON in a text block. Fields that are null or empty are left out.
  • Size. A result is at most 60,000 characters. A longer list is cut and says so in truncated.
  • Errors come back as a tool error with the API’s code and message, for example not_found: No problem with id or slug 'pay-buton' in project 'shop'.
  • What results never contain: what a person typed, replay recordings, page snapshots, ingest keys or tokens.

health

EngineCloud

Answer «what is failing in production?» in one call. Returns the state of the app right now (failing, degraded or ok), the open problems that hit the most people, each told in one plain sentence, any uptime check that is down, pages that got slower, and how fresh the data is. Call this first when the user asks how production is doing, whether a deploy broke something, or what to fix next. To fix one of the problems it returns, call problem with its slug.

Calls: GET /v1/summary

Input

  • since
    string optional

    How far back «now» reaches: a number and a unit (1h, 24h, 7d), or an ISO 8601 timestamp. From 1h to 7d.

    Default: 24h

  • environment
    string optional

    The environment to look at, for example production or development.

    Default: production

Call

JSON
{
  "name": "health",
  "arguments": {}
}

Result

JSON
{
  "project": {
    "id": "prj_5k2m9x4q7w1z",
    "slug": "shop",
    "name": "Shop"
  },
  "environment": "production",
  "generated_at": "2026-03-12T10:35:00Z",
  "data_through": "2026-03-12T10:34:10Z",
  "state": "failing",
  "sentence": "1 new problem in the last 24 hours. Every uptime check is up. 1 page got slower.",
  "problems": {
    "open": 1,
    "new": 1,
    "regressed": 0,
    "top": [
      {
        "id": "prb_8f3k2m9x4q7w",
        "slug": "pay-button-does-nothing",
        "sentence": "3 people tried to pay today and could not. The Pay button does nothing since yesterday's change.",
        "title": "createPayment answered 400 in POST /api/checkout",
        "confidence": "high",
        "people": 3,
        "sessions": 3,
        "first_seen": "2026-03-12T09:12:04Z",
        "last_seen": "2026-03-12T10:31:47Z",
        "first_release": "4f2a91c",
        "regressed": false
      }
    ]
  },
  "uptime": {
    "checks": 1,
    "down": []
  },
  "performance": {
    "slower_pages": [
      {
        "route": "/search",
        "lcp_p75_ms": 3100,
        "previous_lcp_p75_ms": 1900,
        "views": 1840
      }
    ],
    "failing_api": []
  },
  "traffic": {
    "sessions": 498,
    "people": 412,
    "errors": 5
  },
  "alerts": {
    "unacknowledged": 1
  }
}

problems

EngineCloud

List the problems Sypnia found in the app. A problem is a group of failures with one cause (an error, a failing API call), with how many people it hit and since when. Each item has a plain sentence when Sypnia analysed it (sentence), else its technical title (title), plus where in the code and a confidence in the cause. Use it to choose what to work on, to check whether a specific failure is already known (q), or to see what a release broke (release). By default it returns open problems, latest activity first. To fix one, call problem with its slug.

Calls: GET /v1/problems

Input

  • status
    string optional

    open: happening or not yet marked fixed. resolved: marked fixed. ignored: known and accepted.

    Default: open

  • q
    string, up to 120 characters optional

    Words to find in the sentence, the title or the file name, for example checkout or amount_required.

  • release
    string optional

    Only problems seen in this release (the app’s version, usually a short git commit such as 4f2a91c).

  • route
    string optional

    Only problems on this page or API route, for example /checkout or /api/orders/:id.

  • since
    string optional

    Only problems seen after this moment: 24h, 7d, or an ISO 8601 timestamp.

    Default: 30d

  • sort
    string optional

    people puts the problems that hit the most people first. One of: last_seen, first_seen, people, occurrences.

    Default: last_seen

  • environment
    string optional

    The environment to look at.

    Default: production

  • limit
    integer, 1 to 50 optional

    How many problems to return.

    Default: 10

  • cursor
    string optional

    The next_cursor of the previous call, to get the next page.

Call

JSON
{
  "name": "problems",
  "arguments": {
    "status": "open"
  }
}

Result

JSON
{
  "problems": [
    {
      "id": "prb_8f3k2m9x4q7w",
      "slug": "pay-button-does-nothing",
      "status": "open",
      "regressed": false,
      "sentence": "3 people tried to pay today and could not. The Pay button does nothing since yesterday's change.",
      "title": "createPayment answered 400 in POST /api/checkout",
      "confidence": "high",
      "where": {
        "file": "src/checkout/checkout.tsx",
        "line": 142,
        "function": "confirmPayment"
      },
      "route": "/checkout",
      "people": 3,
      "sessions": 3,
      "occurrences": 5,
      "first_seen": "2026-03-12T09:12:04Z",
      "last_seen": "2026-03-12T10:31:47Z",
      "first_release": "4f2a91c",
      "url": "https://app.sypnia.com/shop/problems/pay-button-does-nothing"
    }
  ],
  "counts": {
    "open": 1,
    "resolved": 4,
    "ignored": 1
  },
  "next_cursor": null
}

problem

EngineCloud

Get one problem with its full context for fixing it. Call this before you change any code. It returns: analysis (what happened in one sentence, who it hit, since when, the likely cause with its evidence, what to do, and a confidence level); stack (the frames of the failure in the project’s own files, with the source lines around each; for a failure on the server it is the stack of the error the browser got); breadcrumbs (what the person did just before, click by click); trace (the failing request followed from the browser to the API to the calls behind it); release (the version it started in, with its commits and changed files); impact (how many people and sessions); and sessions to look at. File paths are relative to the repository root. When analysis.status is not_enough_evidence or off, no cause is given: find it yourself from the stack, the breadcrumbs and the release’s changes, and say so to the user. Parts that are missing are listed in missing with the reason.

Calls: GET /v1/problems/{problem}/context

Input

  • id
    string required

    The problem’s slug (for example pay-button-does-nothing) or its id (prb_…), as returned by problems, health or search.

  • session
    string optional

    Build the stack, breadcrumbs and trace from this session’s occurrence instead of the latest one. A session id from sessions.

  • breadcrumbs
    integer, 0 to 100 optional

    How many events before the failure to include.

    Default: 30

  • source_lines
    integer, 0 to 20 optional

    Lines of source around each frame, on each side.

    Default: 5

Call

JSON
{
  "name": "problem",
  "arguments": {
    "id": "pay-button-does-nothing"
  }
}

Result

JSON
{
  "problem": {
    "id": "prb_8f3k2m9x4q7w",
    "slug": "pay-button-does-nothing",
    "kind": "failed_call",
    "status": "open",
    "title": "createPayment answered 400 in POST /api/checkout",
    "route": "/checkout",
    "where": {
      "file": "src/checkout/checkout.tsx",
      "line": 142,
      "function": "confirmPayment"
    },
    "first_seen": "2026-03-12T09:12:04Z",
    "last_seen": "2026-03-12T10:31:47Z",
    "first_release": "4f2a91c",
    "occurrences": 5,
    "people": 3,
    "sessions": 3,
    "url": "https://app.sypnia.com/shop/problems/pay-button-does-nothing",
    "origin": "server"
  },
  "analysis": {
    "status": "ready",
    "sentence": "3 people tried to pay today and could not. The Pay button does nothing since yesterday's change.",
    "who": "3 people. Two of them tried again and gave up.",
    "doing": "Paying for an order, at the last step of checkout.",
    "since": "Yesterday's change, deployed at 18:40. It worked before.",
    "cause": {
      "text": "Release 4f2a91c renamed the cart's total from cart.total to cart.totalCents. The checkout still reads cart.total, so the payment goes out without an amount and the API answers 400.",
      "evidence": [
        {
          "kind": "release_change",
          "ref": "4f2a91c:src/cart/cart.ts",
          "text": "The release renames `total` to `totalCents` in the Cart type."
        },
        {
          "kind": "stack_frame",
          "ref": "src/checkout/checkout.tsx:142",
          "text": "confirmPayment() still passes `cart.total` to createPayment()."
        },
        {
          "kind": "trace",
          "ref": "7c2e9a4b1d8f4e6a9b3c5d7e1f2a4b6c",
          "text": "POST /api/checkout answered 400 with the code `amount_required`."
        },
        {
          "kind": "timing",
          "ref": "4f2a91c",
          "text": "No failure of this kind before the release went live. All 5 failures are on it."
        }
      ]
    },
    "what_to_do": "Read cart.totalCents in confirmPayment(), and show the customer a message when a payment is rejected.",
    "confidence": "high"
  },
  "stack": {
    "origin": "browser",
    "error": {
      "name": "ApiError",
      "message": "POST /api/checkout failed with 400 (amount_required)"
    },
    "symbolicated": true,
    "frames": [
      {
        "function": "createPayment",
        "file": "src/lib/api.ts",
        "line": 27,
        "in_app": true,
        "source": {
          "start_line": 25,
          "lines": [
            "  const res = await fetch('/api/checkout', { method: 'POST', body: JSON.stringify(input) });",
            "  if (!res.ok) {",
            "    throw new ApiError('POST', '/api/checkout', res.status, await errorCode(res));",
            "  }",
            "  return res.json();"
          ]
        }
      },
      {
        "function": "confirmPayment",
        "file": "src/checkout/checkout.tsx",
        "line": 142,
        "in_app": true,
        "source": {
          "start_line": 140,
          "lines": [
            "async function confirmPayment(cart: Cart) {",
            "  setPaying(true);",
            "  const payment = await createPayment({ amount: cart.total, currency: cart.currency });",
            "  router.push(`/orders/${payment.orderId}`);",
            "}"
          ]
        }
      },
      {
        "function": "onClick",
        "file": "src/checkout/pay-button.tsx",
        "line": 18,
        "in_app": true,
        "source": null
      }
    ]
  },
  "breadcrumbs": [
    {
      "offset_ms": 12412,
      "type": "page_view",
      "text": "Opened /checkout (from /cart)"
    },
    {
      "offset_ms": 27090,
      "type": "click",
      "text": "Clicked button «Pay $64.00» in «Your order»"
    },
    {
      "offset_ms": 29310,
      "type": "rage_click",
      "text": "Clicked button «Pay $64.00» 4 more times"
    },
    {
      "offset_ms": 31000,
      "type": "request",
      "text": "POST /api/checkout answered 400 amount_required after 3,870 ms",
      "trace_id": "7c2e9a4b1d8f4e6a9b3c5d7e1f2a4b6c"
    }
  ],
  "trace": {
    "id": "7c2e9a4b1d8f4e6a9b3c5d7e1f2a4b6c",
    "duration_ms": 3870,
    "status_code": 400,
    "spans": [
      {
        "name": "POST /api/checkout",
        "kind": "browser",
        "service": "browser",
        "start_offset_ms": 0,
        "duration_ms": 3870,
        "status": "error"
      },
      {
        "name": "POST /api/checkout",
        "kind": "server",
        "service": "shop-api",
        "start_offset_ms": 14,
        "duration_ms": 3842,
        "status": "error"
      },
      {
        "name": "SELECT carts",
        "kind": "db",
        "service": "shop-api",
        "start_offset_ms": 21,
        "duration_ms": 12,
        "status": "ok"
      },
      {
        "name": "createPayment",
        "kind": "http",
        "service": "shop-api",
        "start_offset_ms": 46,
        "duration_ms": 3790,
        "status": "error",
        "attributes": {
          "peer.service": "payments",
          "http.response.status_code": 400,
          "error.type": "amount_required"
        }
      }
    ]
  },
  "release": {
    "version": "4f2a91c",
    "deployed_at": "2026-03-11T18:40:00Z",
    "previous_version": "9c1e7b2",
    "changes": {
      "commits": [
        {
          "sha": "4f2a91c",
          "message": "Rename cart.total to cart.totalCents"
        }
      ],
      "files": [
        {
          "path": "src/cart/cart.ts",
          "additions": 3,
          "deletions": 3,
          "patch": "@@ -8,7 +8,7 @@ export interface Cart {\n   items: CartItem[];\n   currency: string;\n-  total: number;\n+  totalCents: number;\n }"
        },
        {
          "path": "src/cart/cart-summary.tsx",
          "additions": 2,
          "deletions": 2
        }
      ]
    }
  },
  "impact": {
    "people": 3,
    "sessions": 3,
    "occurrences": 5,
    "people_who_retried": 2,
    "last_success_before": "2026-03-11T17:52:30Z"
  },
  "sessions": [
    {
      "id": "8f3a6c1e9b2d4f07a5c3e1b8d6f4a2c0",
      "started_at": "2026-03-12T10:31:16Z",
      "has_replay": true,
      "occurrences": 1,
      "url": "https://app.sypnia.com/shop/sessions/8f3a6c1e9b2d4f07a5c3e1b8d6f4a2c0?at=31000"
    }
  ],
  "missing": []
}

problem_prompt

EngineCloud

Get Sypnia’s ready-made prompt for fixing one problem, as plain text: what happens, where, why, what to do, how to check, followed by the evidence. It is the same text a person copies from Sypnia into a coding agent. Use it when you want a short brief to follow or to hand to another agent. If you are fixing the problem yourself and want structured data, problem returns the same facts as JSON, with more detail.

Calls: GET /v1/problems/{problem}/prompt

Input

  • id
    string required

    The problem’s slug or id.

Call

JSON
{
  "name": "problem_prompt",
  "arguments": {
    "id": "pay-button-does-nothing"
  }
}

Result

JSON
{
  "analysis_status": "ready",
  "prompt": "Fix a production bug reported by Sypnia.\n\nWhat happens: clicking Pay does nothing. 3 people hit it today (2026-03-12),\n  since release 4f2a91c (deployed 2026-03-11 18:40 UTC).\nWhere: src/checkout/checkout.tsx:142, confirmPayment()\nWhy: that release renamed cart.total to cart.totalCents, but\n  checkout.tsx still reads cart.total. The payment goes out\n  with amount undefined and the API answers 400.\n  Confidence: high.\nDo: read cart.totalCents in confirmPayment(), and show the\n  customer a message when a payment is rejected.\nCheck: session 8f3a6c1e9b2d4f07a5c3e1b8d6f4a2c0 shows the clicks.\n  Add a test that fails when the amount is missing.\n\n… (the evidence and the rules follow; the full text is on the page «The prompt»)"
}

sessions

EngineCloud

List sessions: visits of real people to the app. With problem, returns the sessions that problem affected, with how many times it happened in each and when. Without it, lists recent sessions, by default only those with trouble (an error, a rage click, a dead click or a failed request); filter by person, route or release. Use it to see who was affected and to pick a session to read with session. People are returned as the ids the app gave them, with a name or e-mail only if the app sent those.

Calls: GET /v1/problems/{problem}/sessions or GET /v1/sessions

Input

  • problem
    string optional

    A problem’s slug or id: return the sessions it affected.

  • person
    string optional

    Only sessions of the person with this id. Not with problem.

  • route
    string optional

    Only sessions that visited this page, for example /checkout. Not with problem.

  • release
    string optional

    Only sessions on this release. Not with problem.

  • trouble
    boolean optional

    false also returns sessions where nothing went wrong. Ignored with problem.

    Default: true

  • since
    string optional

    Only sessions after this moment.

    Default: 7d

  • environment
    string optional

    The environment to look at.

    Default: production

  • limit
    integer, 1 to 50 optional

    How many sessions to return.

    Default: 10

  • cursor
    string optional

    The next_cursor of the previous call.

Call

JSON
{
  "name": "sessions",
  "arguments": {
    "problem": "pay-button-does-nothing"
  }
}

Result

JSON
{
  "sessions": [
    {
      "id": "8f3a6c1e9b2d4f07a5c3e1b8d6f4a2c0",
      "started_at": "2026-03-12T10:31:16Z",
      "duration_ms": 42000,
      "person": {
        "id": "u_1042",
        "name": "Ada Park"
      },
      "device": "desktop",
      "browser": "Chrome 134",
      "release": "4f2a91c",
      "errors": 1,
      "rage_clicks": 1,
      "failed_requests": 1,
      "has_replay": true,
      "occurrences": 1,
      "first_occurrence_at": "2026-03-12T10:31:47Z"
    },
    {
      "id": "2b7d9e4f1a6c4830b5d2f7e9c1a3b5d7",
      "started_at": "2026-03-12T09:47:40Z",
      "duration_ms": 82000,
      "person": {
        "id": "u_0977"
      },
      "device": "mobile",
      "browser": "Safari 18",
      "release": "4f2a91c",
      "errors": 2,
      "rage_clicks": 0,
      "failed_requests": 2,
      "has_replay": true,
      "occurrences": 2,
      "first_occurrence_at": "2026-03-12T09:48:11Z"
    },
    {
      "id": "c4e1a8f3d7b24c69a0e5d3b1f9c7a5e3",
      "started_at": "2026-03-12T09:11:26Z",
      "duration_ms": 109000,
      "person": null,
      "device": "desktop",
      "browser": "Chrome 134",
      "release": "4f2a91c",
      "errors": 2,
      "rage_clicks": 0,
      "failed_requests": 2,
      "has_replay": true,
      "occurrences": 2,
      "first_occurrence_at": "2026-03-12T09:12:04Z"
    }
  ],
  "next_cursor": null
}

session

EngineCloud

Read what one person did in one session, as a timeline in plain words: pages opened, things clicked (by their visible label), rage clicks, API calls that failed, errors, forms the browser refused, and custom events. This is the text form of the session’s replay: it tells you the steps to reproduce a failure. It never contains what the person typed, and it does not return video or screenshots. By default only notable events are returned; set notable to false to also see successful API calls and performance measurements. Use around to read the moments before and after a failure.

Calls: GET /v1/sessions/{session}/events

Input

  • id
    string required

    The session id, 32 hex characters, from sessions or from a problem’s sessions.

  • notable
    boolean optional

    true: page views, clicks, rage and dead clicks, errors, failed requests, form failures and custom events. false: everything.

    Default: true

  • around
    timestamp optional

    An ISO 8601 timestamp inside the session. Returns the events around it instead of from the start.

  • limit
    integer, 1 to 500 optional

    How many events to return.

    Default: 100

  • cursor
    string optional

    The next_cursor of the previous call.

Call

JSON
{
  "name": "session",
  "arguments": {
    "id": "8f3a6c1e9b2d4f07a5c3e1b8d6f4a2c0"
  }
}

Result

JSON
{
  "session": {
    "id": "8f3a6c1e9b2d4f07a5c3e1b8d6f4a2c0",
    "started_at": "2026-03-12T10:31:16Z",
    "duration_ms": 42000,
    "person": {
      "id": "u_1042",
      "name": "Ada Park"
    },
    "device": "desktop",
    "browser": "Chrome 134",
    "release": "4f2a91c",
    "has_replay": true
  },
  "events": [
    {
      "offset_ms": 0,
      "type": "page_view",
      "route": "/cart",
      "text": "Opened /cart"
    },
    {
      "offset_ms": 12412,
      "type": "page_view",
      "route": "/checkout",
      "text": "Opened /checkout (from /cart)"
    },
    {
      "offset_ms": 27090,
      "type": "click",
      "route": "/checkout",
      "text": "Clicked button «Pay $64.00» in «Your order»"
    },
    {
      "offset_ms": 29310,
      "type": "rage_click",
      "route": "/checkout",
      "text": "Clicked button «Pay $64.00» 4 more times"
    },
    {
      "offset_ms": 31000,
      "type": "request",
      "route": "/checkout",
      "text": "POST /api/checkout answered 400 amount_required after 3,870 ms",
      "trace_id": "7c2e9a4b1d8f4e6a9b3c5d7e1f2a4b6c"
    },
    {
      "offset_ms": 31004,
      "type": "error",
      "route": "/checkout",
      "text": "ApiError: POST /api/checkout failed with 400 (amount_required)",
      "problem": "pay-button-does-nothing"
    },
    {
      "offset_ms": 42230,
      "type": "page_leave",
      "route": "/checkout",
      "text": "Left /checkout after 30 s"
    }
  ],
  "next_cursor": null
}
EngineCloud

Find things in Sypnia by text when you do not have an id: problems (by words of their sentence, title, file or function), sessions and traces (by id or its first 8 characters), people and accounts (by id, name or e-mail, if the app sent them), releases (by version) and pages (by route). Use it when the user describes a failure in their own words («the coupon error», «what happened to ada@example.com») or pastes an id from somewhere else. Results are grouped by type; follow up with problem, session or sessions.

Calls: GET /v1/search

Input

  • q
    string, up to 120 characters required

    The text to look for.

  • types
    array of string optional

    What to look in. Everything when left out. One of: problems, sessions, people, accounts, traces, releases, pages.

  • since
    string optional

    Only sessions, traces and activity after this moment count.

    Default: 30d

  • environment
    string optional

    The environment to look at.

    Default: production

  • limit
    integer, 1 to 25 optional

    Most results per type.

    Default: 5

Call

JSON
{
  "name": "search",
  "arguments": {
    "q": "ada@example.com"
  }
}

Result

JSON
{
  "q": "ada@example.com",
  "problems": [],
  "sessions": [],
  "people": [
    {
      "id": "u_1042",
      "name": "Ada Park",
      "email": "ada@example.com",
      "account": {
        "id": "acc_77",
        "name": "Corner Café"
      },
      "last_seen": "2026-03-12T10:31:58Z",
      "sessions": 6,
      "trouble_sessions": 1,
      "latest_sessions": [
        {
          "id": "8f3a6c1e9b2d4f07a5c3e1b8d6f4a2c0",
          "started_at": "2026-03-12T10:31:16Z",
          "trouble": true
        }
      ]
    }
  ],
  "accounts": [],
  "traces": [],
  "releases": [],
  "pages": []
}

performance

EngineCloud

See how fast the app is for real visitors and what got worse. Without route it returns the app’s page-load numbers (LCP, INP, CLS and TTFB at the 75th percentile), the pages that got slower against the period before, the slowest pages, API calls that are slow or failing, pages where people rage-click, and how many improvement suggestions are waiting. With route it returns the detail of one page: its numbers, the element behind each one (which image is the slow largest paint, which control is slow to respond), the elements people rage-click or click with no effect, and forms the browser refused. Use it when the user asks why a page is slow, what to optimise, or whether a release made things slower. Times are in milliseconds.

Calls: GET /v1/performance, or with route: GET /v1/performance/pages, /v1/performance/vitals, /v1/clicks and /v1/form-failures

Input

  • route
    string optional

    A page route, for example /search. Leave it out for the whole app.

  • since
    string optional

    The period to measure. It is compared with the period of the same length just before.

    Default: 7d

  • device
    string optional

    Only measurements from this kind of device. All when left out. One of: mobile, tablet, desktop.

  • environment
    string optional

    The environment to look at.

    Default: production

Call

JSON
{
  "name": "performance",
  "arguments": {
    "route": "/search",
    "since": "24h"
  }
}

Result

JSON
{
  "route": "/search",
  "page": {
    "views": 1840,
    "lcp_p75_ms": 3100,
    "previous_lcp_p75_ms": 1900,
    "inp_p75_ms": 180,
    "cls_p75": 0.04,
    "ttfb_p75_ms": 410,
    "poor_share": 0.21
  },
  "elements": {
    "lcp": [
      {
        "selector": "main ul.results > li:nth-child(1) img",
        "kind": "image",
        "samples": 1310,
        "p75_ms": 3200,
        "poor_share": 0.24
      }
    ],
    "inp": [
      {
        "selector": "form.filters select#sort",
        "kind": "pointer",
        "samples": 640,
        "p75_ms": 210,
        "poor_share": 0.03
      }
    ],
    "cls": [
      {
        "selector": "aside.promo",
        "samples": 480,
        "p75": 0.06,
        "poor_share": 0.02
      }
    ]
  },
  "clicks": [
    {
      "selector": "ul.results li .badge",
      "role": "text",
      "label": "Free shipping",
      "clicks": 31,
      "rage_clicks": 2,
      "dead_clicks": 9
    }
  ],
  "form_failures": [],
  "suggestions": [
    {
      "id": "sug_3h7j1k5l9p2o",
      "title": "The search page's first product image loads slowly",
      "status": "new"
    }
  ]
}

uptime

EngineCloud

See whether the app answers from outside. Returns every uptime check (a URL that Sypnia requests on a schedule) with its status (up, down, paused), since when it is down, how the latest check went and the share of checks that passed in the last 24 hours and 30 days. With check it also returns that check’s recent incidents, each with its start, end and reason. Use it when the user asks whether the site is down, or was down, and for how long.

Calls: GET /v1/uptime/checks, and GET /v1/uptime/checks/{check_id}/results with check

Input

  • check
    string optional

    A check’s id (chk_…), to also get its incidents.

  • since
    string optional

    With check: how far back to look for incidents.

    Default: 24h

Call

JSON
{
  "name": "uptime",
  "arguments": {}
}

Result

JSON
{
  "down": 0,
  "checks": [
    {
      "id": "chk_1q9w3e7r5t2y",
      "name": "Shop home",
      "url": "https://shop.example.com/",
      "status": "up",
      "down_since": null,
      "last_checked_at": "2026-03-12T10:34:00Z",
      "last_status_code": 200,
      "last_response_ms": 184,
      "uptime_24h": 100,
      "uptime_30d": 99.98
    }
  ]
}

resolve_problem

EngineCloud

Mark a problem as resolved after its fix is merged or deployed. This is the only tool that changes anything in Sypnia. Call it only when the fix exists in the repository, not when you have merely proposed it, and tell the user you did. It is safe: if the failure happens again Sypnia reopens the problem by itself and alerts. Pass release with the version that carries the fix when you know it, so failures from people still on the old version do not reopen it. Resolving a problem that is already resolved changes nothing.

This tool writes.

Calls: POST /v1/problems/{problem}/resolve

Input

  • id
    string required

    The problem’s slug or id.

  • release
    string, up to 64 characters optional

    The version of the release that carries the fix, usually the short git commit of the fix once deployed.

  • note
    string, up to 500 characters optional

    One line on what fixed it, for whoever reads the problem later: the commit, the pull request, the cause.

Call

JSON
{
  "name": "resolve_problem",
  "arguments": {
    "id": "pay-button-does-nothing",
    "release": "5b3c8d2",
    "note": "Fixed in pull request 128."
  }
}

Result

JSON
{
  "id": "prb_8f3k2m9x4q7w",
  "slug": "pay-button-does-nothing",
  "status": "resolved",
  "resolved_at": "2026-03-12T11:20:44Z",
  "resolved_by": "mcp:Cursor",
  "resolved_in_release": "5b3c8d2",
  "note": "Fixed in pull request 128."
}

Resources

For clients that let a person attach a resource to the conversation. Each returns the same data as a tool.

ResourceWhat it is
sypnia://problems/openThe open problems in production, as problems returns them
sypnia://problem/{id}One problem with its context, as problem returns it. id is the slug or the id
sypnia://problem/{id}/promptThe ready prompt of one problem, as plain text

Prompts

Ready-made requests that MCP clients show as commands (in Claude Code, /mcp__sypnia__fix_problem).

PromptArgumentsWhat it asks the agent to do
whats_failingNoneCheck production with health and report in plain words, without changing code
fix_problemid, optionalRead the problem’s context, confirm the cause in the code, make the smallest fix, add a test, show the change, and call resolve_problem once you confirm the fix is merged or deployed

The full text of fix_problem:

Text
Fix the Sypnia problem «{id}». If no id is given, call `problems` with sort `people` and take the first.
1. Call `problem` to get its context. Read the analysis, the stack and the release's changes.
2. Open the files it names and confirm the cause in the code before changing anything. If the code does not match the analysis, stop and tell me what you found.
3. Make the smallest change that fixes the cause. Do not touch unrelated code.
4. Add a test that fails without the fix and passes with it. Run the tests.
5. Show me the change. Do not merge or deploy.
6. When I confirm the fix is merged or deployed, call `resolve_problem` with a one-line note.

Sections

Start hereWhat Sypnia is, how it works, what is open source and what is the paid cloud, and the two ways to get your first problem in plain words.ConceptsThe ten ideas Sypnia is made of, each defined precisely, from a project and its keys to problems, traces and alerts.SDKsHow your app sends data to Sypnia. One SDK for the browser, one for Node, one for Next.js, and the OpenTelemetry intake for every other backend.The agentHow Sypnia analyses a problem, what it writes, what it sends to a model and what it never does.MCP and skillsConnect your coding agent to Sypnia so it can ask what is failing in production, read what it needs to fix it, and set Sypnia up for you.API referenceThe Sypnia HTTP API: one API for the open-source engine and for Sypnia Cloud, with every endpoint, parameter, response and error.CLI referenceThe sypnia command: how to install it, the variables it reads, what its exit codes mean, and where each command is documented.Self-hostingWhat you run when you host the Sypnia engine yourself, what you need, what it costs, and what you do not get compared with Sypnia Cloud.How the engine is builtThe whole Sypnia engine on one page, from a click in your app to the sentence your coding agent reads, with the design rules and the code that holds them.Sypnia CloudHow the paid service works, from organizations and roles to plans, limits, billing, where your data lives, the GitHub App and support.Privacy and securityWhat Sypnia collects from your app and its users, what it never collects, how masking works, how to delete data, and how the service is protected.LicensingWhich licence covers each part of Sypnia, what you may do with it, how the name may be used, and the terms of a contribution.GuidesStep-by-step recipes for the common ways to use Sypnia, from apps made with AI builders and Next.js on Vercel to Supabase, source maps in CI, alert e-mails from your own engine, and moving from Sentry.ReferenceThe pages you look things up in, from the event schema and the glossary to what is planned, the changelog and the common questions.
↑↓ to move↵ to openEsc to close