Actionable Developer Notifications: Bridging the Gap Between Alerts and Outcomes

Until today, developer notifications were purely informational. You’d get a Slack ping from GitHub Actions or an email from Sentry, and you’d have to manually navigate to OutcomeDev, start a sandbox, and copy-paste the error context just to begin a fix.

That friction ends now.

We are excited to introduce the Developer Notifications API—a high-end integration layer that turns external events into actionable compute.

The "Smart Pivot" Scenario

Imagine an agent is mid-way through a low-priority task: "Update the CSS padding across the dashboard." It's going to take 15 minutes.

Suddenly, Sentry detects a critical crash in production. A script hits our API:

curl -X POST https://outcomedev.com/api/v1/notify \
  -H "X-Outcome-API-Key: otk_live_..." \
  -d '{
    "title": "CRITICAL: Prod Crash",
    "body": "Null pointer in auth.ts:88. Production is down.",
    "metadata": { "taskId": "css-refactor-id" }
  }'

The agent is interrupted mid-turn. It reads the injection, sees the priority, stashes the CSS work, and immediately pivots to fixing the production crash. This is not just a notification; it's a Real-Time Instruction.

From Alert to Action in One Click

Staged Re-entry: The Return Address for Code

Sometimes, an automated fix isn't enough. A human needs to step in. But "stepping in" usually means spending 15 minutes setting up the environment, checking out the branch, and installing dependencies.

Staged Re-entry solves this. By providing a sandboxRefId in the notification metadata, you aren't just telling the human what happened; you are giving them the keys to the exact room where it happened.

When you click "Start Task" on a staged notification, the OutcomeDev UI pre-selects the repository, the task, and the existing sandbox. You are one click away from being live in the terminal session where the build failed. It’s the ultimate context-preservation tool.

When a notification arrives, it’s not just a text bubble. It’s a Seed.

By clicking "Start Task," you aren't just opening the app; you're hydrating our standard Task Form with the precise context of the failure. The repository is pre-selected, the prompt is pre-filled, and even the sandbox choice is optimized based on the notification's metadata.

The Decoupled Compute Lifecycle

We’ve designed this API with total flexibility. You don’t always want a new task for every notification. Sometimes, you want to pivot an existing mission or wake up a cold agent.

Secure by Default

Security shouldn't be an afterthought. You can now issue and revoke developer tokens directly from your Settings > Developer menu.

We use SHA-256 hashing for token storage, ensuring your "Outcome Tokens" (otk_live_...) are never exposed in our database. You see the key exactly once, and after that, we only keep a hashed preview for your reference.

Getting Started

Integrating is as simple as a curl command. Inside your sandboxes, we've already injected OUTCOME_TASK_ID (your current conversation) and OUTCOME_SANDBOX_ID (your durable workspace) into your environment. You can use these to report back to your current mission or target any other task in the system programmatically.

Check out the Notifications API Documentation to see how you can start automating your outcomes today.

The era of passive alerts is over. Welcome to actionable compute.