# Wire MCP health checks into CI

> A ping in your pipeline catches broken keys and endpoint drift before your agents fail silently in production.

- For: Platform engineer
- Category: Agent & Workflow Automation
- MCP tools: ping
- HTML version: https://veridive.com/mcp/use-cases/mcp-health-check-ci

## The problem

The failure mode of agent infrastructure is silence. A revoked API key, a rotated secret that didn't propagate, an endpoint config that drifted in one environment — none of these page anyone, because the agent doesn't crash; it just stops producing, and "the morning brief didn't arrive" gets noticed on Thursday for a Monday breakage.

Every other dependency in your stack gets a health check. The MCP connections your workflows lean on deserve the same one-liner: a scheduled or CI-time reachability probe that turns silent decay into a red build — the difference between infrastructure and hope.

## How veridive MCP helps

**ping** is the probe: a free, single-purpose call that verifies your client can reach veridive and the key is alive. In CI, it runs wherever your agent configs deploy — the same pipeline that ships an MCP config change confirms the connection it configured. On a schedule, it's the heartbeat: a cron'd assistant task that pings and alerts on failure, so key expiry pages someone Monday instead of surfacing Thursday.

The pattern costs one tool call and catches the whole class of silent-decay failures: revoked keys, secret-rotation misses, environment drift, endpoint changes. It's the least glamorous tool in the catalog and the one that makes the rest dependable.

## How to run it

1. **Connect veridive MCP** — One-time setup: add the veridive MCP server to Claude, ChatGPT, Cursor or any MCP client and generate your API key. Takes about two minutes.
2. **Probe on deploy** — Wherever agent configs ship, the pipeline verifies the connection it just configured.

   ```
   Ping veridive and report the result — this is a deployment health check; fail loudly if unreachable.
   ```
3. **Heartbeat on schedule** — A daily scheduled probe — silence in your workflows should always mean 'no findings', never 'dead pipe'.

   ```
   Schedule a daily task: ping veridive; if it fails, alert me with the error — otherwise stay silent.
   ```
4. **Preflight the expensive runs** — Any unattended workflow opens with the probe — the pattern the briefing and budget agents already use.

   ```
   Before the scheduled research run: ping first; on failure, notify and skip the run instead of burning the slot.
   ```

## Problems it solves

- Agent infrastructure fails silently — no crash, just absence, noticed days late.
- Key rotations and secret misses surface as mystery downtime instead of red builds.
- Environment drift (staging works, prod doesn't) hides until the prod run matters.
- Every scheduled workflow inherits the same unverified-connection risk.

## Why it's valuable

- One free call converts silent decay into a paged failure.
- Config deploys self-verify: the pipeline that ships the change confirms it works.
- The heartbeat makes workflow silence trustworthy — 'no news' becomes information.
- Cheapest reliability win in the catalog; every other use case runs on top of it.

## Example prompts

```
Ping veridive as a deploy health check — fail loudly if unreachable.
```

```
Daily scheduled ping; alert on failure, silent on success.
```

```
Preflight the research run with a ping; skip and notify on failure.
```

## FAQ

**Does ping cost anything?**

No — it's a free call, which is what makes it viable as a per-deploy and per-run habit.

**What does ping actually verify?**

Reachability and authentication: your client can reach the veridive server and the key is valid. It's a connection probe, not a full functional test — for that, a cheap free call like library_list makes a good second stage.

**How does this run 'in CI' exactly?**

Via whatever executes MCP calls in your pipeline — a headless assistant step or an MCP-capable runner invoking the tool and asserting success. The pattern matters more than the mechanism: probe where configs change.

## Related use cases

- [Schedule a morning intelligence briefing](https://veridive.com/mcp/use-cases/morning-intel-briefing-agent.md) — A scheduled agent checks the connection, sweeps the last day's uploads in your space, and drops summaries before your first coffee.
- [Give your research agent a spending limit](https://veridive.com/mcp/use-cases/budget-guarded-agent.md) — The agent checks remaining credits before each deep run and degrades to lighter tools when the budget thins.
- [Run one workflow across Claude, ChatGPT and Cursor](https://veridive.com/mcp/use-cases/same-workflow-every-client.md) — MCP means write once: verify each client with ping and reuse the same research workflow wherever you work.

---

Part of the veridive MCP use-case library: https://veridive.com/mcp/use-cases.md
Server docs for agents: https://veridive.com/mcp.md