← Back to all sparks
O

Oh Dear

INFRA · APIS
Velocity5.0

Website monitoring service covering uptime, broken links, certificates and scheduled tasks

The monitoring account is becoming something an agent operates, not just a dashboard someone reads.

uptime-monitoringmcpai-agentsauthorizationstatus-pagesdeveloper-tools
Current state
Oh Dear is an uptime and site-health monitoring service shipping a dated changelog most weeks, with each release mixing new capability, security work and fixes. The MCP server is the thread running through all of it: monitor create, update and delete arrived in May, tag groups and notification routing in June, and the August release extends it across downtime, check summaries, domain intelligence, maintenance windows, check state, status page updates, notification destinations, scheduled tasks, Broken Links, Application Health and account management. Around that sit steady monitoring improvements — consecutive-failure thresholds for sitemap alerts, raw JSON bodies in HTTP checks, better handling of out-of-order scheduled-task pings.
Where it's heading
Two things are being built at once and they depend on each other. The MCP surface keeps widening until an assistant can run the account end to end, and the authorization model keeps being tightened to make that safe — API token scoping, permission checks on monitor-modifying tools, access controls on the new write tools, sessions invalidated on password change. Each release moves both. Underneath, the business plumbing is being tidied in public too: legacy pricing shown on the plan chooser, corrected affiliate terms, an agency-focused pricing structure, AI wizard placeholder monitors excluded from usage limits.
Prediction
Expect the MCP tool surface to keep filling in the remaining corners of the product and the permission model to keep pace with it, and expect the November Slack connection deprecation already being warned about in-app to produce a migration release before then.

Recent moves

  1. 4d ago

    MCP tools extended across maintenance, status pages and account management

    The largest MCP expansion yet, taking the tool surface across downtime, maintenance windows, check state, status pages, notification destinations, scheduled tasks and account management, with access controls strengthened for the new write tools. It extends an agent-operation arc that began with monitor create, update and delete in May rather than starting one.

    View source ↗
  2. 11d ago

    API token scoping and MCP permission checks tightened

    A permissions release: restricted API tokens are held to the monitors they are scoped to, and MCP tools that modify monitors now require matching permissions. This is the safety half of the agent work, and it lands a week before the tool surface widens again.

    View source ↗
  3. 19d ago

    Recovery notifications restored after performance delta alerts

    Three fixes, the notable one being that recovery notifications were never sent after performance delta alerts, leaving teams without an all-clear. Also stops monitors on metered billing being deleted by the monitor limit check.

    View source ↗
  4. 25d ago

    JSON request bodies in HTTP checks; AI monitoring moves to the Laravel AI SDK

    HTTP uptime checks gain raw JSON request bodies with syntax highlighting, and the AI monitoring stack moves from Prism to the official Laravel AI SDK with automated prompt evaluations and validation against private and reserved IPs. Infrastructure work with one real capability attached.

    View source ↗
  5. 1mo ago

    Application Health metadata limit doubled; scheduled-task ping fixes

    Mostly internal: marketing attribution for signups, a doubled Application Health metadata limit, URL canonicalization, Laravel 13 and PHP 8.5 support in the webhooks package, and a fix for scheduled tasks stuck Pending when start and finish pings landed in the same second.

    View source ↗
  6. 1mo ago

    Agency pricing introduced; Forge import moves to scoped API v2 tokens

    Introduces agency-focused pricing and moves the Forge site import to API v2 with scoped tokens, alongside a run of authorization fixes covering team member management, roles and two-factor settings. The pricing change is aimed at the multi-client agencies that run many sites under one account.

    View source ↗