AI & developer tools
Your agent ships the project — it can set up the monitoring too. And inside Probecast, AI does the writing nobody has time for: drafts you review, hypotheses you check. Never a verdict and never sent to a client without you.
For your coding agent
Every tool below speaks to the same API, with the same quotas and client folders as the dashboard — and on Agency, the same audit log. Keys are read-only or read-write — revoke one at any time.
MCP server
Eight tools your agent can call: list_clients, add_monitor, list_monitors, remove_monitor, get_incidents, get_status, get_quota and check_mcp_status. Listed in the official MCP Registry as io.probecast/mcp.
claude mcp add probecast \
-e PROBECAST_API_KEY=pb_live_… -- npx -y @probecast/mcp| Basic | Indie | Pro | Agency |
|---|---|---|---|
| – | 2 | 5 | 15 |
REST API
Create and remove websites, APIs and heartbeats, read incidents and check status. Writes accept an Idempotency-Key, so a retried deploy never creates the same monitor twice.
| Basic | Indie | Pro | Agency |
|---|---|---|---|
| – | 2 | 5 | 15 |
CLI
The same operations from a terminal or a CI step, with no dependencies to install.
npx probecast add --type heartbeat \
--name "nightly backup" --client "Acme"| Basic | Indie | Pro | Agency |
|---|---|---|---|
| – | 2 | 5 | 15 |
llms.txt
A plain-text summary of what Probecast does and how to use the API, for any agent that reads the web before it writes code.
Watch the AI you depend on
A project that calls OpenAI or relies on an MCP server inherits their outages. Both are in the catalog, watched like any other provider.
Model providers
Add the AI APIs a project calls to its client folder; their incidents show up next to yours.
| Basic | Indie | Pro | Agency |
|---|---|---|---|
| 10 | 10 | 60 | all |
MCP servers
Each server is asked to open a protocol session — answering HTTP is not enough to count as up. MCP server status →
| Basic | Indie | Pro | Agency |
|---|---|---|---|
| 10 | 10 | 60 | all |
Where Probecast uses AI
Basic and Indie have no AI. The numbers and alerts their plan includes arrive written from plain templates — anomaly, correlation and prediction alerts are on Agency and webhooks start on Pro.
Monitor setup notes
When you add an API, a short read of what the endpoint likely does and how to watch it.
Briefing
The summary of your projects at 06:00, 12:00 and 18:00 UTC, polished into a few calm sentences.
Weekly digest
The week in one paragraph: what changed, what deserves a look.
Revenue & heatmap insights
A written hypothesis next to the revenue charts and the availability heatmap.
Status page updates
A first draft of the title and text for an incident or maintenance post.
Monthly client report
The narrative of the report, drafted from the month’s data for your sign-off.
Correlation hypotheses
When revenue drops alongside an incident, a short note on the possible link.
Incident post-mortems
A full post-mortem draft from the timeline and the providers involved.
Browser-flow scripts
Describe a login or checkout in plain words; get a script to review and run.
| Where | What it does | Plans |
|---|---|---|
| Monitor setup notes | When you add an API, a short read of what the endpoint likely does and how to watch it. |
|
| Briefing | The summary of your projects at 06:00, 12:00 and 18:00 UTC, polished into a few calm sentences. |
|
| Weekly digest | The week in one paragraph: what changed, what deserves a look. |
|
| Revenue & heatmap insights | A written hypothesis next to the revenue charts and the availability heatmap. |
|
| Status page updates | A first draft of the title and text for an incident or maintenance post. |
|
| Monthly client report | The narrative of the report, drafted from the month’s data for your sign-off. |
|
| Correlation hypotheses | When revenue drops alongside an incident, a short note on the possible link. |
|
| Incident post-mortems | A full post-mortem draft from the timeline and the providers involved. |
|
| Browser-flow scripts | Describe a login or checkout in plain words; get a script to review and run. |
|
How we use it
Every AI text is written as a hypothesis — “the data suggests” — and the numbers it rests on are shown next to it.
Nothing AI-written reaches a client on its own. Reports, status updates and post-mortems wait for you to review them.
When AI is unavailable, you get the plain version or a clear error — never invented filler. Usage has daily limits per account, so a loop can't run up a bill.