The product

    Everything between "it broke" and "it is fixed."

    Six capabilities, one platform, no dashboards to live in. The short version is below; the full detail is in the docs.

    Monitoring

    Everything your production emits, in one place.

    Logs, traces, and metrics over standard OpenTelemetry, no proprietary agent. Auto-instrumentation covers your framework, your database, and your outbound calls, and LLM calls show up as spans in the same trace. Any language OpenTelemetry speaks, Fixter speaks.

    docs: telemetry →
    claude code — fixter onboarding
    > /fixter:onboard
    detected:next.js + postgres
    wiring opentelemetry...done
    telemetry:logs, traces, metrics flowing
    fixter — issues
    duplicates grouped, alerted once:
    TypeError: cannot read "user" of undefinedcheckout-service · 14 events · 1 issue
    StripeCardError: card_declinedpayments · 6 events · 3 users
    anomaly: checkout latency 3x baselineauto-detected · started 14:02

    Issues

    Every issue, caught and grouped. Out of the box.

    Errors and exceptions are captured and deduplicated: the same error is one issue, with counts and affected users. You are alerted when a new issue appears, once, not on every repeat. And because issues live in the same telemetry as your logs, traces, and metrics, each one links straight to its trace. More than exception tracking, and no separate tool to run.

    docs: issues →

    Debugging

    Ask production what happened, in plain language.

    Query your telemetry from Claude Code, Cursor, or any MCP client. Ask what broke, get the answer with evidence. Prefer clicking? The app has live tail and search. Prefer SQL? That works too.

    docs: debugging →
    claude code
    you

    show me errors from checkout in the last hour

    fixter
    47 errors in checkout-service (last 60m)all trace to db timeout in getOrdersstarted 14:23 after deploy #4821
    fixter — alert setup
    analysed your codebase. suggested alerts:
    checkout payment failuresstripe webhook handler
    auth failure spikeauth middleware
    5xx on /api/ordersorders controller
    db query latency > 2spostgres pool

    Alerting

    Rules you write in a sentence. Detectors you never write at all.

    Define alerts in plain language, and Fixter previews each rule against your real history before you save it. Meanwhile its own detectors run from day one: error bursts, anomalies against learned baselines, error budget burn. Alerts reach you in Slack or email, only when they matter.

    docs: alerting →

    Investigations

    The root cause finds you.

    When an alert fires, an AI agent investigates on the spot: it queries your logs and traces and comes back with the root cause, who is affected, and what to do. Pull it into Claude Code with one command and finish the fix, or let Fixter open a ready for review pull request (beta).

    docs: investigations →
    fixter — investigation
    pulling logs (checkout-service, 1h)47 errors found
    tracing request flow...timeout at db query
    checking recent deploys...deploy #4821 (14:23)
    root cause:missing index after migration
    impact:12% of checkout users, 38 min
    fixter — uptime probes
    probes, every 30s from outside:
    GET /api/health200 · 143ms
    POST /api/checkout (staging)200 · 312ms
    mcp handshake + tools/list11 tools · 89ms
    GET /api/orders504 · timeout

    Uptime monitoring

    Know you are down before anyone else does.

    Scheduled probes hit your endpoints from the outside, down to every 30 seconds, with uptime and latency tracked over time. So you catch an outage even when there is no traffic. And it is not just HTTP: Fixter probes your MCP servers the way an agent would call them, handshake, tools and all.

    docs: uptime, API & MCP testing →

    Trust

    Your code stays yours.

    Monitoring and alerting run entirely on the telemetry you send. They never touch your code. Investigations work on telemetry alone too, but grant them read access and they get sharper: the agent sees the code paths behind the failure and hands you a fix that actually fits your codebase. Your repository is cloned into a secure sandbox, used only to solve that problem, and deleted right after.

    Under the hood

    See it on your own production.

    or talk to us