Changelog | Trigger.dev Features, updates, and improvements to Trigger.dev.

That's March.

Full changelog here: trigger.dev/changelog

10 Claude Code Tips You Didn't Know | Trigger.dev Skills teach AI agents how to do specific tasks consistently. Learn how they work, what's inside them, and how agents discover and load them.

3. Plus 10 Claude Code tips most people miss.

--𝚏𝚘𝚛𝚔-𝚜𝚎𝚜𝚜𝚒𝚘𝚗 for context pre-warming. --𝚏𝚛𝚘𝚖-𝚙𝚛 to resume the agent that wrote your PR. ! for inline shell output. 𝙲𝚝𝚛𝚕+𝙶 to compose prompts in your editor.

trigger.dev/blog/10-cla...

How we give every user SQL access to a shared ClickHouse cluster | Trigger.dev A deep dive into TRQL — the SQL-like language behind Query. How we parse, validate, and compile queries to secure, tenant-safe ClickHouse SQL.

2. Deep dive on how TRQL works: the SQL-like language behind Query.

Users write familiar SQL, but the ANTLR grammar physically can't express 𝙸𝙽𝚂𝙴𝚁𝚃 or 𝙳𝙴𝙻𝙴𝚃𝙴. Every query has its tenant filter injected at compile time.

trigger.dev/blog/how-tr...

Why we replaced Node.js with Bun for 5x throughput | Trigger.dev How we took our warm start service from 2,099 req/s on Node + SQLite to 10,700 req/s on a compiled Bun binary, and found a subtle memory leak that only exists in Bun's HTTP model.

New on the blog:

1. We replaced Node.js with Bun in one of our most latency-sensitive services.

2,099 → 10,700 req/s. 5x throughput. Also found a Bun memory leak along the way (which was then swiftly patched by the Bun team).

trigger.dev/blog/firebun

→ v4.4.2 - v4.4.4:

• batch trigger concurrency bumps.
• 𝚜𝚢𝚗𝚌𝚂𝚞𝚙𝚊𝚋𝚊𝚜𝚎𝙴𝚗𝚟𝚅𝚊𝚛𝚜 pulls your Supabase connection strings in so you stop copy-pasting
• Test page generates example payloads with AI
• Dev CLI auto-cancels in-flight runs on exit
• 11 new MCP tools

→ Dashboards

Every project ships with a pre-built one: run volume, success rates, failures, costs, versions. Build custom ones on top with big numbers, charts, and tables. Filters apply across every widget at once.

→ Query.

Ask "what are my p95 durations for the chat task?" and the AI writes the TRQL, runs it on ClickHouse, renders the chart. Or write the SQL yourself.

→ Our @vercel integration is live.

Push code, tasks deploy automatically. Env vars sync both ways. Atomic deployments gate Vercel's promotion until your tasks are ready, so your app never goes live with a mismatched task version.

No 𝚝𝚛𝚒𝚐𝚐𝚎𝚛.𝚍𝚎𝚟 𝚍𝚎𝚙𝚕𝚘𝚢, no CI/CD workflow.

In case you missed it, here are some of the things we shipped in March:

→ Native @vercel integration
→ Query & Dashboards (ask in English, get SQL)
→ 5x throughput from a Bun rewrite
→ New MCP tools, TTL defaults, 𝚜𝚢𝚗𝚌𝚂𝚞𝚙𝚊𝚋𝚊𝚜𝚎𝙴𝚗𝚟𝚅𝚊𝚛𝚜

Full recap ↓

GitHub agent automation: Hookdeck, Trigger.dev, and Claude Build AI-powered GitHub agent automation (PR reviews, issue labels, Slack summaries) with durable Trigger.dev tasks behind Hookdeck: GitHub Source Authentication, transforms, retries, and two routing patterns—task router vs connection routing at the edge.

HookDeck wrote the full walkthrough: both routing patterns, deployment scripts, the Hookdeck transform code, everything.

Fork it and have AI-powered GitHub automation running in ~15 minutes: hookdeck.com/webhooks/pl...

2 routing architectures:

1 Hookdeck connection → router task dispatches child tasks via 𝚝𝚊𝚜𝚔𝚜.𝚝𝚛𝚒𝚐𝚐𝚎𝚛() with 𝚒𝚍𝚎𝚖𝚙𝚘𝚝𝚎𝚗𝚌𝚢𝙺𝚎𝚢: ${𝚍𝚎𝚕𝚒𝚟𝚎𝚛𝚢𝙸𝚍}-${𝚝𝚊𝚜𝚔𝙸𝚍}. Simple.

Or: 3 connections filtering on 𝚇-𝙶𝚒𝚝𝙷𝚞𝚋-𝙴𝚟𝚎𝚗𝚝 header → tasks directly. Per-event retry policies, independent pause/replay, separate delivery

The idempotency pattern is the interesting part.

A Hookdeck transform extracts GitHub's 𝚇-𝙶𝚒𝚝𝙷𝚞𝚋-𝙳𝚎𝚕𝚒𝚟𝚎𝚛𝚢 header into the payload. That becomes the idempotency key. Hookdeck retries the delivery? Same key. No duplicate task runs. No duplicate PR comments. Zero application code for dedup.

3 tasks, each in its own container:

𝚑𝚊𝚗𝚍𝚕𝚎-𝚙𝚛: fetches the diff, calls Claude, posts a single review comment (HTML anchor 𝚊𝚒-𝚛𝚎𝚟𝚒𝚎𝚠-𝚜𝚞𝚖𝚖𝚊𝚛𝚢 prevents duplicates on re-runs)

𝚑𝚊𝚗𝚍𝚕𝚎-𝚒𝚜𝚜𝚞𝚎: Claude classifies, parses JSON labels, applies via GitHub API

𝚑𝚊𝚗𝚍𝚕𝚎-𝚙𝚞𝚜𝚑: Claude summarizes commits in Slack

GitHub → Hookdeck (HMAC verification, payload reshape, retries) → Trigger REST API→ durable task container → Claude + GitHub API + Slack

No custom webhook server. No Express route parsing 𝚇-𝙷𝚞𝚋-𝚂𝚒𝚐𝚗𝚊𝚝𝚞𝚛𝚎-𝟸𝟻𝟼. No retry logic

GitHub sends a webhook. 60 seconds later, Claude has reviewed your PR diff, labeled your issue, and posted a deploy summary to Slack.

Built with HookDeck + Trigger.dev + Claude ↓

AI Engineer Europe 2026: Apr 8-10, London The first flagship AI Engineer Europe event! Bringing our large multitrack technical AI conference to Europe for over 1000 AI Engineers. Apr 8-10, 2026 • London, UK

www.ai.engineer/europe

We're heading to AI Engineer Europe in London!

April 8-10 | QEII Centre, London UK | Booth G6

Come and chat with us If you're building AI agents, workflow automation, or background jobs. We'll be doing live demos, sharing what we're building, and handing out swag. See you there!

10 Claude Code Tips You Didn't Know | Trigger.dev Skills teach AI agents how to do specific tasks consistently. Learn how they work, what's inside them, and how agents discover and load them.

The gap between "I use Claude Code" and "I orchestrate Claude Code" is wide and getting wider.

These aren't novelty tricks. They're the building blocks for treating AI as a managed fleet of specialised workers, not a chat partner.

Bookmark this. You'll come back to it.

tgr.dev/2TACXLp

10/ Headless CI/CD with hard budget caps

Putting an autonomous agent in a pipeline without boundaries is terrifying.

-p for non-interactive mode
--max-turns to prevent infinite loops
--max-budget-usd as a financial circuit breaker

9/ Dynamic multi-agent orchestration with --agents

You don't need to hardcode subagent definitions in markdown files.

Pass a JSON definition to --agents for session-scoped specialists.

8/ Context compaction with Esc+Esc

Long debugging sessions fill the context window with dead weight. Every failed attempt adds tokens that degrade performance.

Double-tap Esc, select a message midway through, choose "Summarise from here."

Everything before that point is preserved perfectly.

7/ Structured JSON output with --json-schema

When you're building automation pipelines, conversational output is useless.

Combine -p, --output-format json, and --json-schema to turn Claude into a strictly typed function.

You define the shape. The model is constrained to produce exactly that.

6/ Parallel agents with --worktree

This one is genuinely transformative.

Multiple Claude sessions against the same repo = race conditions, trampled file edits, chaos.

--worktree uses native git worktree to give each agent a completely isolated working directory.

5/ Effort levels for cost control

Not every task needs deep reasoning. Boilerplate generation doesn't deserve the same compute as debugging a race condition.

The /model command now has 4 effort tiers: Low, Medium, High, Max

For CI/CD, set it programmatically:

export CLAUDE_CODE_EFFORT_LEVEL=low

4/ Inline shell execution with !

Prefix any input with ! to bypass the LLM entirely and run it in your shell.

The key detail: stdout and stderr get automatically injected into the context window.

Run ! npm test, see the failure appear, type "fix it", and Claude already has the full error context.

3/ Escape the REPL with Ctrl+G

The single-line input is actively hostile to serious prompt engineering.

Ctrl+G opens your $EDITOR (Vim, VS Code, whatever). Suddenly you have macros, syntax highlighting, multiple cursors, and proper multi-line editing.

2/ Code review loops with --from-pr

You wrote the code three days ago. Reviewer left comments this morning. Now you need to mentally reconstruct everything before responding.

Instead:

claude --from-pr 447

The agent comes back with full awareness of its original implementation decisions.

1/ Session forking with --fork-session

Stop rebuilding context from scratch every time.

Create a "master session" loaded with 40k+ tokens of architecture docs and project goals. Then fork it for each new feature.

It's git branch, but for your LLM context window.

Most people use Claude Code like a fancy chatbot in their terminal.

Meanwhile, power users are running parallel agents across git worktrees, forking LLM sessions like git branches, and capping autonomous CI/CD runs with hard budget limits.

10 advanced patterns that change everything 🧵

Vercel integration | Trigger.dev Connect your Vercel project to Trigger.dev and never run a manual deploy command again. Automatic deploys, env var sync, and atomic deployments

Read more in our changelog:

trigger.dev/changelog/v...