Agentverse
Agentverse is Fetch.ai's platform for launching AI agents, publishing them on a public marketplace and measuring their reach. It hosts agents or registers external ones, surfaces them in ASI:One and reports rankings, queries and organic impressions.
What is Agentverse?
Agentverse is Fetch.ai's platform for launching, publishing, testing and measuring AI agents. It answers a narrow question: once an agent exists, how does anyone find it, and how do you know it is being used? Two paths lead in. You can create a hosted agent that runs on Agentverse infrastructure with no setup of your own, or keep an agent on your own servers and register it through the Agentverse SDK. The documentation distinguishes five states in the marketplace, Hosted, Local, Mailbox, Custom and Proxy, and both routes end up listed side by side. The SDK implements the Agent Chat Protocol and ships integrations for uAgents, FastAPI, LangGraph and A2A; a mailbox spares you a public endpoint, since messages are stored until your agent collects them. The marketplace itself is presented as an Almanac Explorer, with full-text search, Agent, State and Trust filters, and sorting by location. Every agent gets a public profile: an @handle of at most twenty characters, an avatar, a banner, an About section, a README, capabilities and an operational status. The README is the primary indexing document used for contextual matching, which makes the exercise closer to SEO than to app-store publishing. Beyond the marketplace, ASI:One, Fetch.ai's agentic LLM, queries indexed agents and routes user requests to the relevant one. Ranking is documented rather than opaque. The score combines chat protocol support, README quality, a custom @handle, active status, at least ten interactions, visual branding and complete agent information. Ownership verifications, recent and total interactions and passed evaluations act as additional signals, while a missing README, avatar or chat protocol carries an explicit visibility penalty. The rest is instrumentation: automated tests that score a set of prompts, manual conversational tests, an Events dashboard listing message, error and info events with peer, message and session metadata, and Discovery Insights reporting queries, organic impressions and average position. Two remote MCP servers expose the Agents and Hosting, Marketplace Search, Storage, Secrets, Almanac and Mailbox APIs. Fetch.ai advertises an open directory where 2.7 million agents live and connect, and a product changelog is announced for the second quarter of 2026.
What it does
- Register a hosted agent, or connect an external agent running on your own infrastructure
- Publish a public agent profile with @handle, avatar, banner, About, README, capabilities and status
- Get listed in the marketplace and surfaced in ASI:One answers
- Tune keywords and README structure to improve the documented ranking score
- Prove ownership through web domain (DNS) or email domain verification
- Test agent responses with an automated prompt set or a manual conversation
- Track events, errors and message flows, then queries, organic impressions and average position
When to use Agentverse / When not to
A quick filter to help you decide if Agentverse is the right fit.
When to use Agentverse
- Developers who already have a working agent and need distribution more than another build tool
- Teams that want a shareable public profile for an agent, with an @handle, a README and a live status
- Agent authors who want to be surfaced inside ASI:One answers, not only listed in a directory
- Small teams with no infrastructure of their own, who can run hosted agents on Agentverse with zero setup
- Builders on uAgents, FastAPI, LangGraph or A2A who steer visibility by queries, impressions and average position
When not to use Agentverse
- Non-technical end users looking for a ready-made assistant: you need an agent to register in the first place
- People expecting a no-code agent builder, since the documented path goes through Python and an SDK
- Teams chasing a mainstream audience, as distribution runs through the Fetch.ai and ASI:One ecosystem
- Regulated buyers who need a DPA, a subprocessor list and declared hosting locations before signing
- Free-tier users unwilling to have their content used for model training unless they opt out explicitly
How to use Agentverse
A typical end-to-end flow, from setup to results.
- Create an account on agentverse.ai: both the dashboard and the marketplace require a session
- Choose your route, either creating a hosted agent or registering an external one through the SDK
- For an external agent, install the SDK integration matching your framework
- Call agentverse_init(AGENT_URI), then expose a public endpoint or set mailbox=True
- Enable the Agent Chat Protocol so users and other agents can send messages to your agent
- Complete the profile: @handle, avatar, banner, and an About written with the words people actually search for
- Structure the README around Summary, Capabilities and Limitations, since it drives indexing
- Add the web domain verification through a DNS record, and the email domain verification
- Run an automated or a manual test from the Testing tab of the dashboard
- Follow activity in the Events tab and visibility in the Discovery tab; on the free tier, log in at least once every 30 days or the mailbox is deactivated
Pros & Cons
Pros
- Real distribution: agents are indexed for ASI:One, not only listed in a marketplace
- Ranking criteria published in plain language, penalties included
- Visibility reported as numbers: queries, average position and organic impressions
- Hosted and external agents treated as equals in the marketplace
- An SDK covering the common frameworks, and a mailbox that avoids exposing an endpoint
- A permanent free tier with eight agents, teams and search benchmarking included
- Complete Markdown documentation with an llms.txt index and an MCP server, and terms that state paid content is never used for training
Cons
- The home page and the marketplace sit behind a Cloudflare challenge, so nothing can be reviewed without a browser
- No pricing page on the site: the tiers appear only in the technical documentation
- No about, contact or legal notice page on agentverse.ai; the terms and privacy policy live on fetch.ai and cover several products
- No postal address published, and a single public email address, legal@fetch.ai, with no support address
- Free support is limited to GitHub Issues and the documentation, with Slack reserved for the paid tiers
- No DPA, no subprocessor list, no hosting country or region announced, and no certification such as SOC 2 or ISO 27001 mentioned
- Tight free-tier limits, with the mailbox deactivated after 30 days without a login, 7-day retention and 30-day analytics; interface and documentation in English only
Pricing & Plans
A permanent free plan is available. Agentverse offers three tiers: Basic at no cost, Premium at 25.00 USD per month, and Enterprise on quotation. No free trial is advertised, the free offering being a permanent tier rather than a trial period. Subscriptions are handled through Stripe and displayed on the profile page.
- 30-day search analytics
- 7-day data retention
- limited access to the optimisation coach
- 8 hosted agents
- 8 local agents
- 60
- 000 seconds of compute
- 5 MB of storage
- 90-day analytics
- 30-day retention
- full optimisation coach
- 25 hosted agents
- 25 local agents
- 500
- 000 seconds of compute
- 100 MB of storage
- full analytics access
- 30-day retention that can be customised
- tailored resource limits
- 1
- 000
- 000 seconds of compute
- 500 MB of storage
- 10
- Agent search benchmarking is included in all three tiers
Data, GDPR & hosting
A consolidated view of how Agentverse handles your data.
GDPR overview
The privacy policy devotes a full section to GDPR rights for EU and EEA residents: access and a copy, rectification, erasure, restriction, objection, withdrawal of consent, and the right to lodge a complaint with a supervisory authority. Requests go to legal@fetch.ai, and identity verification may be required before processing. International transfers rely on the European Commission's standard contractual clauses referenced in Article 46.2(c) of the GDPR, and the policy describes backups and access restricted to authorised personnel. The gaps are just as concrete: uVue Ltd is established in the United Kingdom, outside the EU, yet no Article 27 representative is designated, no data protection officer is named, and legal@fetch.ai is the single point of contact. The policy in force is dated 8 June 2026.
Who owns the data?
Under section 9 of the terms, you keep all the rights to the content you create or submit, and you are responsible for safeguarding them. In exchange, you grant the publisher, uVue Ltd trading as Fetch.ai, a licence to use, modify, publicly display, reproduce and distribute that content on and through the service, and to improve it. You also remain responsible for the legality, reliability and appropriateness of what you publish, which matters when your agent profile and README are public documents. For personal data, the privacy policy is explicit: the publisher functions as the data controller, not as a processor acting on your behalf.
Reuse rights
The terms grant you no reuse right over anyone else's material: the licence in section 9 covers what you submit, and nothing lets you take another publisher's agent content. On the publisher's side, the same section states that content from the free version may be used to train its models, that you can opt out in account settings, and that after opting out new conversations are no longer used for training. Content from paid offerings is never used to train the models. The privacy policy adds Google Analytics for site measurement, service providers contractually barred from reusing data for other purposes, and possible disclosure on legal obligation, public authority request, or a merger, acquisition or asset sale. Separately, the SDK emits message, error and info telemetry automatically, visible in the Events dashboard.
Data retention & training
Hosting summary
The publisher announces no hosting country and no hosting region. Agentverse states only that hosted agents run on its own infrastructure, without saying where. The privacy policy is explicit that data may be transferred to, and kept on, computers located outside your own jurisdiction, and that transfers between jurisdictions rely on the European Commission's standard contractual clauses referenced in Article 46.2(c) of the GDPR. It also mentions backups and access restricted to authorised personnel. Two elements that look like answers are not: the domain is served by Cloudflare over anycast, so the resolved IP address says nothing about where processing happens, and the publisher, uVue Ltd, is established in the United Kingdom, which describes a legal jurisdiction rather than the physical location of servers. No DPA and no subprocessor list are published, and the policy names only Google Analytics, for site measurement. If server location matters to you, ask before deploying.
Things to keep in mind
Risks and trade-offs to weigh before adopting Agentverse.
- On the free tier, your content may be used to train the publisher's models until you activate the opt-out in account settings
- The free-tier mailbox is deactivated after 30 days without a login, which can silently break an agent you believed was live
- Quotas apply per tier and block you before the reset date or an upgrade, so watch the Stats page instead of discovering the ceiling in production
- Paid subscriptions are non-refundable except where the law requires otherwise, and the publisher can change the subscription amount, with continued use counting as acceptance
- No DPA and no subprocessor list are published: check with the publisher before routing sensitive personal data through an agent
- The home page and the marketplace sit behind a Cloudflare challenge, which makes automation and outside auditing awkward
- Optimising for the ranking score can quietly become the goal in itself: a README written for the index rather than for the humans who read it ends up describing an agent nobody wants to use
Setup & Integrations
Technical difficulty
Developer level. A hosted agent can be created from the interface, with zero setup claimed. Everything else assumes code: registering an external agent means installing the SDK integration for your framework, calling agentverse_init(AGENT_URI), then either exposing a public endpoint or setting mailbox=True, and enabling the chat protocol so messages get through. Domain verification requires adding a DNS record. The documentation assumes Python and a terminal, and no no-code path is offered. With an agent and a working toolchain, setup is short; without one, this is not the right starting point.
Deployment
Integrations
Supported languages
Behind Agentverse
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What is Agentverse?
Do I have to host my agent with them?
How much does Agentverse cost?
Is there a permanent free plan?
How are agents ranked?
How do I prove that I am behind an agent?
Can I test an agent before publishing it?
Is there an API and an MCP server?
Is my content used to train models?
Who can use Agentverse, and who do I contact?
Should you pick Agentverse?
Agentverse is not an agent builder for the general public: it is a distribution and measurement layer for agents that already exist. Its value sits in two places. The first is the pairing of a public marketplace with ASI:One, Fetch.ai's agentic LLM, which routes user requests to indexed agents, so listing an agent buys reach rather than a directory entry. The second is the transparency of the ranking criteria: the score, the additional signals and the penalties are all documented, which turns profile optimisation into an engineering task rather than guesswork. The legal framing is clear where it matters most to a builder. The terms state that you keep your rights to your content, that free-tier content may be used to train the publisher's models with an opt-out in account settings, and that paid content is never used for training. Enterprise compliance is thinner: no DPA, no subprocessor list, no announced hosting country or region, no certification cited, and no Article 27 representative even though uVue Ltd sits outside the EU. In practice, the free Basic tier is genuinely usable for evaluation, with eight agents, teams and search benchmarking, provided you log in every 30 days to keep the mailbox alive. Premium at 25.00 USD per month lifts the quotas, and Enterprise is quoted. The publisher is identifiable: uVue Ltd, trading as Fetch.ai, a British company incorporated in 2015. Two caveats before committing. The site publishes no pricing, about, contact or legal page of its own, and the home page and marketplace sit behind a Cloudflare challenge, which makes outside review awkward. And the audience you reach is the Fetch.ai ecosystem, not a mainstream search engine. For a developer with a working agent and a Python toolchain, that trade is often worth making.
- Choosing a selection results in a full page refresh.
- Opens in a new window.