LoginSign Up
← Back to all sparks
T

Trigger.dev

INFRA · APIS
Velocity3.1

Trigger.dev is reshaping itself into the runtime layer for AI and agent workflows.

job-orchestrationai-agentsmcpdeveloper-toolsinput-streamsanalytics
◆Current state
Trigger.dev shipped a steady run of v4.4.x releases (4.4.0 through 4.4.5) with a clear theme stack: input streams for bidirectional communication into running tasks, a Query & Dashboards surface with SQL analytics over your run data, deeper MCP server tooling (11 new tools in 4.4.4), an error-tracking dashboard, and a Vercel integration with automatic deploys. Operational polish — task-level TTL defaults, run replay detection, headless CLI flag, longer API key rotation grace — fills the gaps.
◆Where it's heading
Two patterns dominate. First: AI-and-agent specialization — input streams are exactly the primitive an agent runtime needs to feed planning state into a long-running task, and the MCP tooling is the public surface agents call to use Trigger as a job runner. Second: self-service operations — auto-cancelling dev runs on CLI exit, default TTLs, the new dashboards — a sign the team is pulling teams off scripts and onto Trigger as a managed platform.
◆Prediction
The next minor (4.5) likely formalizes the agent-runtime story — typed agent invocation contracts on top of input streams, broader MCP coverage, and probably an explicit "agent task" task type. Expect more integrations following the Vercel template (likely Netlify and Render next) since those are the deploy targets where Trigger needs to be invisible.

◆Recent moves

  1. 1mo ago

    Trigger.dev v4.4.5

    v4.4.5 lands operational polish: ctx.run.isReplay so tasks can detect they're being replayed, a --no-browser CLI flag for CI and SSH sessions, and a 24-hour grace window when rotating API keys. Quiet release that smooths the sharp edges users hit in production.

    View source ↗
  2. 2mo ago

    Input streams: send data into running tasks

    ⚡ SPARK

    A dedicated changelog post — not just a release note — introduces input streams: typed bidirectional channels into a running task, with four receiving patterns from suspending (which frees compute) to non-blocking. This is a new primitive, not a tweak.

    View source ↗
  3. 2mo ago

    Trigger.dev v4.4.4

    v4.4.4 adds 11 new MCP server tools and major improvements to the existing MCP surface, plus task-level TTL defaults. The MCP investment is sustained — this is the third release in a row touching that surface — confirming Trigger sees MCP as table stakes for being agent-callable infrastructure.

    View source ↗
  4. 3mo ago

    Vercel integration

    A first-class Vercel integration with automatic deploys, env-var sync, and atomic deployments removes a friction point most users hit early: keeping Trigger task code synchronized with frontend deploys. The first deploy-target integration; expect more.

    View source ↗
  5. 3mo ago

    Trigger.dev v4.4.3

    v4.4.3 ships an error-tracking dashboard that consolidates failures across runs, plus Supabase environment-variable sync and dev-run auto-cancel when the CLI exits. Together these reduce the operational tax of running Trigger long-term.

    View source ↗
  6. 3mo ago

    Query & Dashboards: analytics for your Trigger.dev data

    Query & Dashboards lands: SQL analytics over Trigger.dev run data, with chart and table dashboards on top. Pulls reporting in-platform instead of forcing teams to ETL run data into a separate analytics stack.

    View source ↗