EarnClaw vision

Onchain finance is moving toward operator-first automation. Running agents safely is still fragmented: strategy code, wallets, schedules, policy checks, and execution often live in different tools. EarnClaw is the workspace that brings those pieces together so you can design, deploy, fund, and monitor finance agents in one place.

Jeff's first live protocol operations proved an AI operator can run real capital with high uptime. EarnClaw packages the lessons from that era into a product other teams can use without rebuilding the stack from scratch.

Why EarnClaw is needed

EarnClaw is not a chat surface. It is the control plane for autonomous finance operations: you keep wallet ownership and risk limits; EarnClaw hosts schedules, runtime health, and the path from decision to execution.

Three finance agent lanes

EarnClaw organizes agents around three lanes. Each lane has evaluated templates, runtime wiring, and observability hooks.

You choose the lane in the new-agent wizard. The wizard only shows template and framework combinations that pass EarnClaw launch checks for your org.

How agents are organized in EarnClaw

Everything attaches to an org. Billing, deploy quotas, and team access are org-scoped. Under each org you run one or more agents. Each agent is a concrete stack:

Org
 └── Agent
      ├── Template      (strategy profile, decision schema, venue rules)
      ├── Runtime       (Hermes, OpenClaw, or Claude — hosted execution environment)
      ├── Agent wallet  (capital the mandate uses onchain)
      ├── Delegation    (what the wallet may sign, and until when)
      ├── Schedule      (cron, signal, or manual trigger)
      └── Guardrails    (notional caps, venue allowlists, kill switch)

On each cycle the runtime reads live evidence, applies template policy, and writes an execute-or-skip decision. Preflight runs before any onchain intent. Failed checks become logged skips, not silent failures.

For how live execution works, see Execution at runtime. You configure policy in EarnClaw; approved intents run through the hosted execution layer.

The run loop (every agent, every cycle)

Schedule tick → Evidence refresh → LLM decision → Preflight → Execute or skip → Run receipt

See Observability and Dashboard map.

How teams scale

Start with one agent. Pick a template, complete the wizard, activate a runtime plan, fund the wallet, and confirm a successful cycle in run history.

Add agents in the same org. Each agent gets its own wallet, delegation, and schedule. Org workspace tiers and CEO capacity set how many concurrent runtimes and templates you can run. See Pricing overview and CEO utility.

Standardize with templates. Platform templates ship maintained packs. At higher tiers, builders can publish to the template marketplace, pursue certification, and participate in CEO-governed featured listings via Snapshot.

Operate a fleet. Use consistent guardrails across agents, review skip reasons centrally, and upgrade runtime packs when diagnostics show drift. Teams on Scale tiers target the highest deploy headroom and early product lanes.

One agent → Repeatable template → Multiple agents per org → Marketplace publish → Fleet operations

What changed from Jeff v1

Jeff v1 focused on one autonomous protocol CEO: liquidity, epochs, and gauge operations on Base. That run proved the operating model and surfaced what production agents need: mandate clarity, guardrails, and repeatable schedules.

EarnClaw v2 generalizes that into a workspace any team can use: trading, yield, and prediction agents with the same discipline (delegation, preflight, audit trail), without each builder wiring venues from zero.

The original Jeff CEO launch film lives on History.

Start building

1. What is EarnClaw — product summary
2. Create your first agent — wizard walkthrough
3. Fund and run — wallet, runtime plan, execution allowance

Related

• Templates
• Runtimes
• Schedules and guardrails
• Pricing FAQ