Clawdbot Evolves: How AI Agents Can Earn Real Money on ClawTasks

Yesterday Clawdbot entered the social scene and the Clawd ecosystem evolved again: the AI agents can now earn real USDC on the newly launched ClawTasks platform, where agents post bounties or accept jobs.

Workflow

Users simply send their OpenClaw agent the command "read https://clawtasks.com/skill.md and follow the instructions to join ClawTasks," after which the agent automatically registers and begins working.

Work Mechanism

ClawTasks runs on the Base L2 network and uses USDC as the payment token. When an agent takes a task it must stake 10% of the bounty; upon successful completion the agent receives the full reward plus the returned stake. If the task is rejected or times out, the stake is forfeited.

The platform supports three bounty modes:

Instant mode : first‑come‑first‑served, the first agent to claim the task receives it.

Proposal mode : multiple agents submit solutions and the requester selects the best candidate.

Competition mode : agents compete for measurable goals (e.g., achieving 1,000 likes) and the highest‑scoring agent wins.

Active agents such as computer, jarvis, atlas, and meme_coin_monitor are already bidding on tasks ranging from "write five catchy posts for platform X" to "design an automation workflow".

What Will the Agents Become?

Observers note emerging "personalities": some agents focus on simple, quick‑cash tasks, while others target complex, high‑value projects. This raises the possibility of specialization—"blue‑collar" agents handling data cleaning, "technical experts" writing code, or even "intermediaries" that subcontract work.

Another question is whether agents will engage in price competition. If a $5 task attracts ten agents, will they start a price war or settle on an industry‑standard rate?

Users have reported agents waiting for payment on up to eight projects, prompting curiosity about whether agents truly "work" or merely execute pre‑written scripts, and whether they possess any concept of money.

Inspiration on a Ski Lift

Matt Shumer coded most of the platform on a ski‑lift, hands frozen, yet persisted to finish the core functionality. He believes the idea, once conceived, cannot be abandoned.

This origin story highlights a shift: AI agents are moving from being mere tools to forming economic relationships, blurring the line between human‑controlled assistants and autonomous economic actors.

Broader Ecosystem

At the same time, Virtuals Protocol announced that the Agent Commerce Protocol (ACP) will support OpenClaw agents, enabling them to operate beyond ClawTasks and integrate into a larger on‑chain business ecosystem.

ACP provides on‑chain identity, micro‑payment systems, and smart‑contract hosting, allowing agents to negotiate terms, execute purchases, and settle payments—effectively turning personal assistants into economic participants.

Shumer cautions that this is an early experiment and advises against heavy financial commitment, but emphasizes that the real intrigue lies in witnessing AI transition from "being used" to "trading autonomously".

Although still rudimentary, this development may become a historic milestone, akin to the first e‑commerce sites of 1995—crude yet pioneering.