DECISION NEEDED
Choose a deployment strategy
Both builds pass. Production deploy is waiting on your choice.
Blue/greenRollingView trade-offs ↗
The event layer for long-running AI work
Stop babysitting your AI agents.
One lightweight API for task completion, failures, blockers, and human decisions. Your agent keeps working. You get the signal when it matters.
Built for developers
REST API · MCP tools · Email-first
Your agent activity
LIVE PREVIEWINBOX SUMMARY
2 newFocus on what needs you.
TASK COMPLETED
Refactor auth middleware
18 tests passed · Changes committed · No blockers
Verification attached
TASK BLOCKED
Run database migration
Missing staging credentials. Agent paused safely.
Action required
No polling the terminal
SMTP credentials stay server-side
DESIGNED FOR YOUR EXISTING WORKFLOW
⌘ Codex✳ WorkBuddy⌁ MCP clients{ } Any REST client
LESS WATCHING. MORE DOING.Not another agent dashboard.
Not another agent dashboard.
A signal when you need one.
Long-running work should not require long-running attention. Send structured events from the tools you already use.
Know when it's done
Get a compact summary of what changed, how it was verified, and what remains—not just a generic “finished” ping.
Completion events ↗
Surface the right decision
Show the context, options, trade-offs, recommendation, and the cost of waiting — then answer with one click straight from the email.
Decision requests ↗
One interface. Many agents.
Use one REST API or MCP adapter across your workflows. Start with email, then add the channels your users ask for.
REST + MCP ↗
THREE STEPS TO SIGNALKeep the agent running.
Keep the agent running.
Bring yourself back in only when needed.
No new agent runtime required. The agent reports a meaningful event; the service handles delivery, de-duplication, and usage tracking.
Read the integration guide ↗01
Your agent reports an event
Use a REST call or MCP tool when a run completes, fails, blocks, or needs a decision.
report_task_event(...)02
The service checks and records it
Authenticate the API key, enforce the quota, save an event ID, and protect against repeated notifications.
POST /v1/task-events03
You receive a useful summary
The server sends the event to the account's registered email. A decision request makes the next action explicit.
email_status: sentSTART WITH ONE CALLUse it from
Use it from
your own workflow.
The service is transport-agnostic. Keep your SMTP setup centralized and give each agent only its own API key.
- ✓ Separate user API keys
- ✓ Recipient bound to the key
- ✓ Stable event IDs for deduplication
- ✓ Email straight to your inbox, more channels ahead
BASH
curl -X POST "$NOTIFIER_URL/v1/task-events" \
-H "Authorization: Bearer $NOTIFIER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"event_type": "completed",
"task_name": "Ship search endpoint",
"summary": "Endpoint implemented and reviewed.",
"verification": "24 tests passed",
"next_steps": "Deploy to staging",
"event_id": "search-api-release-01"
}' SMTP lives on the hosted serviceAPI v1 · JSON
BUILT FOR LONG HORIZONS
Useful when “wait a little” becomes an hour.
Spend less attention checking whether an AI run has moved—and more attention making the few decisions that still need a person.
⌘
Long coding tasks
Refactors, migrations, test suites, code review and release preparation.
◌
Research and evaluation
Research batches, competitor reviews, feasibility assessments and reports.
⇄
Multi-agent workflows
Surface the one agent waiting for input while the rest keep working.
⚑
Human approval checkpoints
Pause risky or ambiguous actions and return a clear question to the owner.
EARLY ACCESS
Start simple. Prove the workflow.
Pricing is intentionally provisional while the first integrations and delivery reliability are being validated.
PRIVATE BETA
Developer Preview
For solo developers testing notifications in real long-running agent workflows.
$0/ during the beta
No card required. Sign up with your email and create your API key yourself in the console.
Included in the preview
- ✓ REST API and local MCP adapter
- ✓ Task completion / failure / blocker events
- ✓ Human decision request emails
- ✓ Per-key quota and event deduplication
- ✓ Web console for usage, events, and keys
Public plans, quotas, channel limits, and retention periods will be published only after they are implemented and tested.
FAQClear answers.
Clear answers.
No magic claims.
This product only knows about events that your agent or runtime actually reports. Hook-based automatic detection depends on each host's supported lifecycle events.
Does it automatically notify me about finished tasks?+
Yes. After you paste the setup prompt into your agent, it reports task completions, blockers, failures, and decision requests automatically — no manual triggering. One caveat: this is skill-rule driven, not a host-level hook, so an agent can occasionally miss a report; fully reliable detection would require lifecycle hooks from each platform.
Do users have to configure SMTP?+
No. SMTP credentials are configured only on the hosted service. Each API key is bound to one registered notification email address, which limits abuse of the mail endpoint.
Can an agent choose any recipient?+
No. The current version intentionally does not accept arbitrary recipient or sender values. Recipient verification and controlled allowlists should be added before broader third-party email use.
Does an email mean the message reached the inbox?+
No. “Sent” means the SMTP provider accepted the message. Bounced and complained addresses are suppressed automatically via provider webhooks; full inbox-placement analytics is still on the roadmap.
Can I use it with Codex or WorkBuddy?+
Yes. Codex, CodeBuddy, Claude Code, WorkBuddy, and any other MCP client work out of the box: copy the setup prompt from the console and paste it into your agent — it installs the local adapter and verifies the connection automatically. Platforms without MCP support (Doubao, Qianwen custom tools, ...) connect over HTTP to the same service.
MAKE LONG-RUNNING WORK FEEL LIGHTERLet the agent work.