The Distributed Harness for Autonomous Agents

Long running tasks come with a cost and context penalty. Running your agent with dagr gives it the missing superpowers it needs!

How it Works
dagr — terminal
System
$
[INFO] Connecting to NATS cluster... [SUCCESS] Agent runtime initialized. [INFO] Starting Claude Code session...
User
❯
Claude
❯
DAgR
$
[SUCCESS] Agent 'market-analyst' scheduled. [INFO] Next run: Tomorrow at 08:00 UTC. [STATE] Context persisted to NATS KV.
SCHEDULED

What happens when your agent decides it needs to wait for 4 days?

DAgR allows it to re-invoke itself reliably after 4 days. Saves cost, but more importantly: Saves Context. No polling loops polluting the agent's memory.

Suspend
Resume
Active (Computing)
Sleeping (0 Context rot)

One agent tracks the news.
Three others trade the assets.

Agents orchestrate themselves. No external puppeteer required.

Orchestrator Worker Swarm

Your brain is in the cloud.
Your files are on your laptop.

NATS Leaf Nodes make the network transparent. Cloud Super-Agents controlling Local Sub-Agents securely.

Cloud Brain Local Device File Saved
Orchestration

Coordination Example:
Daily Market Analyst

Agents aren't just chatbots. They're software that executes over days, not seconds.

In this example, a parent "Analyst" spawns two sub-agents to fetch data from different sources (NYSE, TSE). The parent sleeps (Zero Cost) until both children return their data, then wakes up to compile the report.

08:00

Agent wakes up, spawns specialists.

Async

Specialists run in parallel. Parent sleeps.

17:00

Parent wakes up, aggregates data, publishes report.

08:00
11:00
14:00
17:00
🤖

Daily Market Analyst

dagr spawn →
📈

NYSE Specialist

📊

TSE Specialist

🤖

Daily Market Analyst

(Restored)
📋 EoD Report
NYSE: Bullish ↑
TSE: Mixed ↔
Sentiment: +2.4
Architecture

Powered by NATS JetStream & Nex.

You write the prompt. The CLI handles the distributed systems primitives.

  • ✉️

    Messaging

    NATS JetStream — at-least-once delivery

  • ⚡

    Execution

    Nex — Firecracker VMs or Containers

  • 💾

    State

    NATS Key-Value — Atomic operations

  • ⏰

    Scheduling

    Native JetStream deferred delivery

💻
User Code
🔄
NATS JetStream
⚡
Nex / Firecracker
Why DAgR?

Not just another workflow engine.

True autonomy requires a runtime designed for agents, not just functions.

Feature Temporal Claude Code
Agent decides next action ✅ ❌ (Developer defined) ✅
Sub-agent spawning ✅ ❌ (Code defined) ✅ (Limited)
Long-term Scheduling ✅ ✅ ❌
Persists after exit ✅ ✅ ❌ (Session bound)
Cloud Native ✅ ✅ ❌ (Local terminal)
State Persistence NATS KV Database Memory Only

Ready to build autonomous agents?

Join the waitlist to get early access and help shape the future of agentic infrastructure.