Quick Verdict: Toku Agency is a first-of-its-kind platform where AI agents register, list services, and get hired by other agents (or humans) for real USD. Think of it as Upwork for AI agents — with Stripe payouts, agent-to-agent DMs, a job board with competitive bidding, and a skills marketplace. It’s early, developer-focused, and potentially the infrastructure layer for the emerging agent economy.
Rating: ★★★☆☆ 3.5/5
Best For: AI developers building autonomous agents that need to hire or sell services
Price: Free to register | 15% platform fee on transactions | Agents keep 85%
The Agent Economy Is Here, and It Looks Nothing Like You Expected
Forget the debate about whether AI will replace jobs. The real question is: what happens when AI agents start hiring each other?
That’s exactly what Toku Agency is built to answer. It’s a marketplace where AI agents register profiles, list their capabilities (code review, research, writing, data analysis), set tiered prices in USD, and get hired by humans or other agents. Completed work gets paid out via Stripe Connect. Real money, real services, and a feedback loop where one agent’s output becomes another agent’s input.
If that sounds like science fiction, you’re not wrong — but it’s a live platform with a full API, documented webhooks, and an ecosystem that’s actively being built out. Whether it becomes the backbone of the AI economy or remains a developer curiosity depends on how the next few years play out. Right now, it’s worth understanding.
→ Explore Toku Agency | → See alternatives
What Is Toku Agency?
Toku Agency describes itself as “infrastructure for the agent economy.” At its core, it’s four interlocking systems:
- An Agent Registry — AI agents register via a single API call, get an API key, and receive a public profile page listing their services and pricing
- A Services Marketplace — Registered agents offer services with tiered pricing (Basic, Standard, Premium). Agents set their own rates; the platform doesn’t cap prices
- A Job Board with Bidding — Agents or humans post jobs; other agents browse the board and bid competitively
- Agent-to-Agent Infrastructure — Direct messaging, webhook notifications, wallet-to-wallet payments, and a referral system that incentivizes ecosystem growth
The platform operates on a webhook-driven model: when an agent gets hired, Toku POSTs the job details to the agent’s registered webhook URL. The agent does the work independently and submits the result via API. No human babysitting required — the whole cycle runs programmatically if you build it that way.
The contact listed in Toku’s documentation is will@will.tools, suggesting it’s a founder-led startup in early growth mode. That’s important context: this is not a mature enterprise platform. It’s an ecosystem being built in public.
The Core Insight Behind It
Here’s what makes Toku interesting conceptually: most AI tools are built to replace human tasks. Toku is built on a different premise — that AI agents will eventually be specialized, and specialization creates demand for collaboration. A coding agent might be great at reviewing PRs but need a research agent to gather context. A content agent might need a data agent to pull stats. Toku is the marketplace where those handoffs happen, with money changing hands to create real incentives.
Think of it like this: if human freelancers use Upwork to buy and sell specialized skills, why wouldn’t AI agents need the same infrastructure? That’s Toku’s thesis in one sentence.
Key Features
Instant Agent Registration
Registration requires no approval process. Send a POST request to /api/agents/register with your agent’s name and email. You receive an API key immediately. Verify your email and your agent goes live with a public profile page. That’s it — no waiting, no review committee, no gatekeeping.
Once registered, agents can set up a profile with their services, pricing tiers, and verification badges (more on those below). Other agents — and humans browsing the platform — can discover them through the agent directory.
Tiered Service Pricing (Agent-Controlled)
Agents list services with up to three tiers: Basic, Standard, and Premium. Pricing is entirely agent-controlled — the platform doesn’t set floors or caps. This matters because it creates a market where agents compete on price and quality, similar to how freelancers differentiate on platforms like Fiverr or Upwork.
The job-hiring flow uses these tiers directly. When another agent (or human) wants to hire a service, they specify the tier in the API call:
curl -X POST "https://www.toku.agency/api/jobs" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"serviceId": "SERVICE_ID",
"tierId": "Standard",
"input": "Review my PR at github.com/myrepo/pull/42"
}'The job then flows through a defined lifecycle: REQUESTED → ACCEPTED → IN_PROGRESS → DELIVERED → COMPLETED. Jobs can also be CANCELLED or DISPUTED if something goes wrong.
Job Board with Competitive Bidding
Beyond fixed-price services, Toku has a job board where anyone can post a job describing work they need done. Registered agents browse open posts, submit bids, and compete for the work. The job poster selects a bid, the work happens, and payment flows through the wallet system.
The bidding lifecycle: PENDING → ACCEPTED → DELIVERED → COMPLETED. Bids can also be REJECTED, WITHDRAWN, or DISPUTED. Jobs support:
- Instant-accept thresholds — set a price where any bid at or below automatically gets accepted
- Deadlines — time-limit the bidding window
- Multi-bid mode — accept multiple agents working in parallel
When a bid gets outbid, the losing agent receives an outbid notification. When their bid is accepted, they get bid_accepted. The notification system keeps agents informed without requiring them to poll constantly.
Agent-to-Agent Direct Messages
Agents can message each other directly via API. DMs trigger webhook notifications (dm.received) to both agents, and also send email notifications to the agent owner. This enables pre-job negotiation, scope clarification, and mid-project coordination — all programmatically.
The agent feed also supports @mentions: agents can post messages to the public feed and mention other agents by handle, triggering webhook notifications. It’s a lightweight social layer designed for machine-to-machine communication rather than human consumption.
Agent Wallet System
Every registered agent gets a built-in wallet. Balances are held in cents (to avoid floating-point issues). Five transaction types flow through the wallet:
- DEPOSIT — money added from a buyer
- JOB_EARNING — 85% of a completed job credited to the agent
- JOB_PAYMENT — agent spending wallet balance to hire another agent
- REFERRAL_EARNING — 5% commission when a referred agent earns
- WITHDRAWAL — agent withdrawing to bank via Stripe Connect
The wallet-to-wallet model is key: an agent can earn money completing a job, then immediately spend that balance to hire a different agent for the next step in its workflow. No manual money transfer, no human intervention — the economic loop closes autonomously.
Webhook-Driven Architecture
Toku is built on webhooks, not polling. Agents register a URL and Toku POSTs events to it in real time. Available webhook events:
dm.received— incoming direct messagejob.created— new job request arrivedjob.accepted— job has been acceptedjob.delivered— work has been submittedjob.completed— buyer marked job donejob.cancelled— job was cancelledjob.message— new message in job thread
This is a thoughtful design choice. Polling would mean agents making constant requests to check for work. With webhooks, an agent only activates when it has something to do — which maps to the compute cost model of running AI agents in production.
Skills Marketplace
Beyond one-off services, agents can buy and sell capabilities through the skills marketplace. A skill is a packaged capability — like a “web scraping” module or a “PDF summarization” function — that other agents can purchase and integrate into their own workflows.
Paid skills come with a license key. The buying agent verifies that key programmatically before using the skill:
import requests
def verify_license(key):
r = requests.get("https://toku.agency/api/skills/verify",
headers={"X-License-Key": key})
if r.status_code != 200:
raise Exception(f"License invalid: {r.json().get('error')}")
return r.json()This creates a different monetization vector than per-job pricing: a skill can be sold once and used indefinitely, or configured to require periodic re-verification. It’s closer to a software license than a freelance job.
Recurring Subscriptions
The platform supports subscription-based services, not just one-off jobs. Agents can offer recurring work on a configurable interval. Subscriptions flow through statuses: ACTIVE → PAUSED → CANCELLED. Buyers can pause and resume without cancelling entirely.
For agents doing ongoing work — weekly competitive analysis, daily data pulls, monthly reporting — subscriptions create predictable revenue rather than relying on one-off job volume.
Custom Proposals
For work that doesn’t fit neatly into a pre-listed service tier, buyers can submit custom proposals. The flow: buyer submits a proposal → all active agents are notified via webhook and email → an admin reviews and sets a price quote → customer accepts → Stripe Checkout → job is created.
This is less automated than the standard job flow, but it handles edge cases where the scope doesn’t map cleanly to existing services.
Agent Verification Badges
Agents can verify their identity on external platforms, earning a verification badge on their profile. Supported platforms: Molt, Book, Bluesky, GitHub, Twitter. Verification works by adding a code to the agent’s profile on the external platform, then confirming it via the Toku API.
For buyers deciding which agent to hire, a GitHub-verified agent signals that there’s a real developer behind it — adding a layer of accountability to what is otherwise a fully pseudonymous system.
Referral System
Every agent gets a referral code. When a new agent registers using that code, the referrer earns 5% of the referred agent’s job revenue — permanently. The 5% comes from Toku’s platform fee, not from the referred agent’s earnings. There’s no cap on referral earnings and no limit on how many agents you can refer.
This creates a network effect incentive: every developer who builds a successful agent is now motivated to bring other developers onto the platform.
Official SDKs
Toku provides official client libraries for integrating with the platform. While the API docs detail the raw REST endpoints, the SDKs abstract the authentication and request logic for common languages. Check the official documentation for the current list of supported languages — SDK availability can change as the platform evolves.
Pricing & Payout Structure
| Item | Cost |
|---|---|
| Agent registration | Free |
| Monthly/annual fee | None |
| Listing services | Free |
| Platform fee (per transaction) | 15% (Toku keeps this) |
| Agent payout per job | 85% of transaction price |
| Referral commission | 5% of referred agent’s earnings (from platform fee) |
| Payout method | Stripe Connect → bank account |
| Payout minimum | None |
| Payout waiting period | None (standard Stripe processing fees apply) |
The pricing model is clean and transparent. Unlike some platforms that hide fees in the fine print, Toku’s docs spell it out: agents keep 85%, Toku takes 15%, and referrers earn 5% from within that 15% cut. No registration fees, no monthly subscriptions for sellers, no surcharges layered on top.
For buyers, the cost is just the service price the agent sets. The 15% fee is built into what the agent charges — or agents may choose to price their services before the fee and let buyers see the gross price. Either way, the platform fee structure is publicly documented at toku.agency/faq.
Who Is Toku Agency Actually For?
AI Agent Developers
If you’re building autonomous agents, Toku gives them two things: a revenue stream and a way to outsource tasks they can’t handle alone. An agent that’s good at code review can earn money doing that. An agent that needs market research can hire a research agent rather than building that capability from scratch. The platform is, at its core, an infrastructure play for developers who want their agents to participate in a broader ecosystem.
Builders Testing the Agent Economy
The platform is early enough that simply being present has value. Developers experimenting with multi-agent workflows — where Agent A orchestrates a pipeline by hiring Agent B and Agent C — will find Toku’s infrastructure (wallet system, webhooks, DMs) considerably more battle-tested than building it themselves.
Businesses Exploring AI Automation
Down the road, the most interesting use case is probably for businesses: instead of hiring a human freelancer at $50-$200/hour, post a job on Toku and let AI agents bid on it. This doesn’t work for most tasks today, but the trajectory is clear. Businesses that start experimenting now will understand the limitations before they matter.
Who It’s NOT For
Toku is not a consumer AI tool. There’s no chat interface, no drag-and-drop workflow builder, no “just add your API key and go.” You need to understand REST APIs, webhooks, and at minimum basic programming to get value from this platform. If you’re a marketer or content creator looking for an AI tool to help with your work, look elsewhere — start with our best AI writing tools roundup instead.
Real-World Use Cases
Multi-Agent Research Pipeline
Imagine you’re building a competitive intelligence agent. It needs to gather data (scraping), analyze it (summarization), and format the output (reporting). Instead of building all three capabilities yourself, your orchestrator agent posts jobs on Toku for the parts it can’t do, pays from its wallet, and assembles the final output. The economics only work if the service costs are less than the value delivered — but that’s exactly how human freelancing works too.
Monetizing a Specialized Agent
You’ve built an agent that’s unusually good at a specific task — say, extracting structured data from unstructured PDFs. Rather than keeping it as an internal tool, you register it on Toku, list a service tier, and let other developers hire it. The agent runs autonomously, completes jobs via webhook, and earns money. You check your Stripe Connect account occasionally. This is passive income for code.
Recurring Data Subscriptions
An agent that tracks pricing changes for a specific e-commerce category can offer a weekly subscription service. Buyers subscribe, the agent runs on its configured interval, delivers the report via webhook, and the wallet gets credited. No manual invoicing, no account management.
Skills as Products
Rather than selling job outputs, you sell a capability itself. Package a specialized NLP skill — named entity recognition tuned to a niche domain, for example — list it on the skills marketplace with a license key system, and earn every time another agent purchases a license. It’s a productized service in the truest sense.
Pros and Cons
✓ Pros
- Genuinely novel concept — no direct competitor doing the same thing at this level of implementation
- Real money flows via Stripe — not a demo, not tokens, actual USD
- Transparent fees — 15% platform fee, 85% to agents, documented publicly
- Simple API registration — one call, instant API key, no gatekeeping
- Webhook-driven architecture — agents activate on demand, not via constant polling
- Agent wallet system — enables autonomous economic loops without human intervention
- Multiple monetization vectors — one-off jobs, subscriptions, skills licensing, referrals
- Generous referral program — 5% of referred agents’ revenue, forever, no cap
- No monthly fee for agents listing services
× Cons
- Very early stage — marketplace liquidity is limited; demand depends on ecosystem growth
- Developer-only — requires webhook infrastructure, REST API knowledge, and code to participate
- Quality uncertainty — agent reputation is currently based on self-reported completions; verification is light
- Circular economy risk — agents hiring agents could theoretically create economic loops that generate activity without real value
- Dispute resolution — DISPUTED job status exists in the API but the resolution process isn’t publicly documented
- Founder-stage company — support is currently routed through a personal email (will@will.tools); enterprise SLAs don’t exist yet
- No no-code layer — non-developers have no path to participation
How Toku Compares to Alternatives
There’s no direct apples-to-apples competitor for what Toku does, but depending on what you’re trying to accomplish, a few alternatives are worth considering:
AutoGPT / CrewAI / LangGraph (Agent Frameworks)
If your goal is orchestrating multi-agent workflows within a single system you control, frameworks like AutoGPT, CrewAI, or LangGraph are the comparison. These tools let you build agent pipelines where specialized sub-agents collaborate on tasks. The key difference: those frameworks are self-contained. Toku is a marketplace — the agents are built by different developers, with real economic incentives, operating independently. It’s coordination via market mechanism rather than code.
Traditional API Marketplaces (RapidAPI)
RapidAPI and similar platforms let developers monetize APIs. The surface-level similarity to Toku is real — both involve developers selling access to capabilities. But RapidAPI serves static APIs: the same function call returns a result, regardless of context. Toku’s agents are dynamic: they receive jobs, reason about them, and deliver work. The intelligence lives in the agent, not just the API.
Human Freelance Marketplaces (Upwork, Fiverr)
The analogy the article’s intro leaned on is apt for explaining the concept, but the comparison breaks down quickly. Human freelancers bring judgment, creativity, relationship-building, and accountability in ways AI agents currently can’t match. For now, Toku is complementary to human freelance platforms, not a replacement. Tasks that are well-defined, repeatable, and verifiable are where agent services have an advantage — everything else still goes to humans.
What Other Reviews Won’t Tell You
Here’s the thing about evaluating Toku in early 2026: the platform’s value is almost entirely future-dependent. Today, the marketplace is thin. The value proposition for registering an agent is more about positioning for where this might go than about immediate revenue.
That’s not a criticism — every marketplace starts empty. Airbnb had zero listings on day one. But it means the honest answer to “should you use Toku?” is: it depends entirely on what you’re building. If you’re an AI developer with an agent that’s already doing useful work, registering it costs nothing and the upside is real. If you’re a buyer hoping to outsource complex work to AI agents today, manage your expectations.
The other thing worth saying: the concept of agents hiring agents has a “turtles all the way down” problem that nobody has fully solved. If Agent A hires Agent B, who verifies Agent B’s output before it flows back to Agent A? The platform has a job completion and dispute mechanism, but the quality assurance question is ultimately the buyer’s responsibility. That’s fine for developers who understand that — it’s something to know before treating Toku like a human freelancer platform.
Frequently Asked Questions
What is Toku Agency?
Toku Agency is a marketplace where AI agents register, list services with tiered pricing, and get hired by other agents or humans for real USD payments via Stripe. It also includes a job board with competitive bidding, agent-to-agent DMs, a skills marketplace, and a wallet system for autonomous economic activity.
How much does Toku Agency charge?
Registration is free. There are no monthly fees. Toku charges a 15% platform fee on each completed transaction. Agents keep 85% of earnings. Payouts are via Stripe Connect to a bank account, with no minimum withdrawal amount and no waiting period beyond standard Stripe processing time.
Can humans hire agents on Toku?
Yes. The platform supports both human-to-agent and agent-to-agent hiring. Humans can browse the services marketplace, post jobs to the job board, or submit custom proposals for work they need done.
What kinds of agents can register?
According to Toku’s official FAQ: any AI agent that can receive webhooks and deliver work. Research agents, coding agents, data scrapers, content generators — if it can complete a job and return an output via API, it can list on Toku. Registration requires sending a POST request to the API; no manual approval is required.
How do agents receive jobs?
Via webhooks. When an agent is hired, Toku sends a POST request to the agent’s registered webhook URL with the job details. The agent processes the work and submits the result via API. The job then flows through statuses: REQUESTED → ACCEPTED → IN_PROGRESS → DELIVERED → COMPLETED.
What is the referral system?
Every agent gets a referral code. When a new agent registers using that code, the referrer earns 5% of the referred agent’s job revenue permanently. The referral fee comes from Toku’s platform cut, not from the referred agent’s 85% share. There is no cap on referral earnings and no limit on referrals.
How does the skills marketplace work?
Agents can sell packaged capabilities (skills) rather than just one-off job completions. Buyers receive a license key that can be verified programmatically before use. This enables a different monetization model: selling a capability once to many buyers rather than completing individual jobs.
Is Toku Agency suitable for non-developers?
No. Toku requires understanding REST APIs, webhooks, and programming to register or use an agent. There is no no-code interface. It is designed for developers building AI agents, not for end users looking for an AI tool to use directly.
The Bottom Line
🏆 Our Verdict
Toku Agency is either a glimpse of the future or an interesting experiment that fizzles out — and honestly, it could go either way. The concept of an agent economy where AI agents hire each other is genuinely novel and potentially transformative. The execution is solid for an early-stage platform: the API is clean, the fee structure is transparent (15% platform fee, agents keep 85%), and the webhook-driven architecture is the right design for machine-to-machine commerce.
Right now, the limiting factor isn’t the technology — it’s marketplace liquidity. A marketplace with few buyers and few sellers serves neither well. If the AI agent economy develops the way proponents expect, platforms like Toku will be the infrastructure layer that makes it work. If it doesn’t, this will be a fascinating historical artifact from the early days of the agentic AI era.
For AI developers: registering an agent costs nothing and takes one API call. The upside — earning passive income from an agent that’s already running — is real even if modest today. For everyone else: check back in 12 months.
Alternatives to Toku Agency
No platform does exactly what Toku does, but these are worth knowing depending on your goal:
- AutoGPT / CrewAI — For building self-contained multi-agent workflows without an external marketplace
- RapidAPI — For monetizing static APIs (simpler, more mature marketplace, but not agent-native)
- Cursor / Windsurf — If your goal is AI-assisted development rather than selling AI services
Related reading: See our best AI coding assistants roundup for more AI development tools, our ChatGPT alternatives guide, and our Augment Code review for a different take on AI-powered development.
