convex-sentinel
get-convex/agent-skills
Capture production errors in your Convex deployment with built-in redaction and optional AI-driven triage.
What is convex-sentinel?
Set up Sentinel to automatically capture server function failures, client crashes, and scale signals into a table in your own Convex deployment. Data never leaves your infrastructure and is redacted at write time by default. Use it to monitor production health and optionally trigger AI-assisted fix PRs for recurring errors.
- Captures server function failures, client JS/React crashes, OCC errors, and scale signals
- Redacts sensitive data at write time with default-deny rules for secrets and patterns
- Stores all error data in your own Convex deployment—no third-party transmission
- Monitors errors via the monitor's `prod_error` event for real-time reactions
- Optionally classifies errors and generates fix PRs via AI-runner for recurring non-transient issues
How to install convex-sentinel
npx skills add https://github.com/get-convex/agent-skills --skill convex-sentinel- A deployed Convex cloud application
- @convex-dev/sentinel package installed
- Convex CLI for reading and querying error data
How to use convex-sentinel
- 1.Install the component by adding `app.use(sentinel)` to `convex/convex.config.ts`
- 2.Wire the client SDK: add a React error boundary and set up `window.onerror`/`unhandledrejection` handlers plus breadcrumbs
- 3.Review redaction settings (default-deny on secret key names and patterns; customize as needed)
- 4.Query recent errors using `convex data` or `run-once-query` in the CLI
- 5.(Optional) Enable the self-healing cron to auto-classify errors and trigger AI-runner for fix PR generation
Use cases
- Monitor production health and error trends in a deployed Convex app
- Automatically capture and redact client-side crashes and server failures
- Trigger custom workflows when new errors occur via the monitor event system
- Use AI triage to classify recurring errors and auto-generate pull requests for fixes
- Track OCC conflicts and scale signals to identify performance bottlenecks
- Backend engineers deploying Convex applications to production
- Teams needing production error monitoring without external error tracking services
- Developers who want AI-assisted error triage and automated fix generation
- Organizations with strict data residency or security requirements
convex-sentinel FAQ
All error data is stored in a table in your own Convex deployment. It never leaves your infrastructure or is sent to third parties.
No. Redaction runs at write time and is mandatory by default, using a default-deny policy on secret key names and value patterns.
Yes. You can sample errors and set caps to control volume and associated costs.
Yes, capturing production errors requires a deployed cloud Convex app, though installation itself works anonymously.
When enabled, the triage cron classifies each error and hands recurring non-transient ones to ai-runner, which can open a fix PR automatically.
Full instructions (SKILL.md)
Source of truth, from get-convex/agent-skills.
name: convex-sentinel description: "Set up Sentinel production error capture in your own Convex deployment."
<!-- GENERATED from convex-agents content/capabilities/sentinel.json — do not edit by hand. -->Capture production errors in your own deployment
Install @convex-dev/sentinel to capture production errors (server function failures, client JS/React crashes, OCC and scale signals) into a table in the user's OWN deployment, redacted at write time, then react to new ones. Data never leaves the user's deployment.
Workflow
- Install the component:
app.use(sentinel)inconvex/convex.config.ts. - Wire the client SDK: a React error boundary plus
window.onerror/unhandledrejectionand breadcrumbs. - Redaction runs at write time and is on by default (default-deny on secret key names and value patterns).
- Read recent errors with the Convex CLI (
convex data,run-once-query); react to new ones via the monitor'sprod_errorevent. - Optionally enable the self-healing cron:
triageclassifies each error and, for recurring non-transient ones, hands it to ai-runner to open a fix PR.
Rules
- Redaction is mandatory and on by default — never store raw secrets; the agent's reads reach the model provider.
- Data stays in the user's deployment; never send it to a third party.
- Sample and cap to control volume and cost.
- Capturing PROD errors needs a deployed cloud app; install works anonymously.
Related skills
More from get-convex/agent-skills and the wider catalog.
convex-setup-auth
Set up secure authentication in Convex with user management and access control.
convex-suggest
Suggest the right Convex component when you detect hand-rolled patterns it already solves.
convex-test
Generate convex-test tests for Convex functions using vitest and an in-memory backend.
convex-verify
Verify Convex features work by seeding data, driving as multiple users, and asserting authorization and behavior.
cargo
Router and entry point for the Cargo CLI skill bundle — load first for multi-domain tasks or when unsure which skill to use.
cargo-ai
Build and configure AI agents in Cargo with prompts, RAG knowledge, MCP servers, and memory management.