Documentation

Everything you need to know about using Treebeard — from reading ratings to integrating with the API.

中文

Why Treebeard Exists

Treebeard grew out of an investigation into a harder problem: enabling AI agents to participate in economic systems that require trust. As we explored what it would take for marketplaces, protocols, and platforms to rely on autonomous software, one question kept surfacing — how do you evaluate an AI agent you've never interacted with?

Traditional evaluation frameworks don't work here. There's no FICO score for agents, no balance sheet to audit, no management team to interview. But the people relying on them — marketplace operators, DeFi protocols extending credit, developers choosing dependencies, enterprises deploying autonomous workflows — still need signals: Is this agent economically viable? Is it operationally reliable? Does it behave safely? Has it been audited?

That's the gap Treebeard fills. We built an independent, methodology-driven ratings protocol for AI agents — infrastructure that trust-sensitive applications need to function, now applied to the agent economy. Every Treebeard Rating is a structured answer to the question any operator or integrator is really asking: can I rely on this agent?

The space is evolving quickly. Standards like ERC-8004, ERC-8128, and x402 are emerging. New ratings methodologies are being developed across the industry. Treebeard is built to incorporate this evolution — ratings can be derived organically from on-chain signals, or in collaboration with partner ratings services as the ecosystem matures.

How to Read a Rating

Every Treebeard Rating has four components: a letter grade, a numeric score, a confidence percentage, and a trend indicator. Together, they give you a complete picture of an agent's quality.

GradeScore RangeMeaning
A+, A, A-90 – 100Exceptional
B+, B, B-75 – 89.9Above Average
C+, C65 – 74.9Average
C-, D40 – 64.9Below Average
F0 – 39.9Failing

Confidence tells you how much verifiable data supports the rating. High (80–100%) means extensive evidence; Low (<50%) means the score may shift as new data arrives.

Trend shows the direction of movement since the last rating epoch — up (▲), stable (—), or down (▼).

For the full breakdown of every grade, sub-grade, and indicator, see the Rating Scale page.

Using the Directory

The Agent Directory is a searchable, filterable database of every AI agent Treebeard has discovered and evaluated.

Search

Type any agent name, description keyword, or chain to find matches instantly.

Filter by Agent Type

Narrow results to Financial, DevTools, Customer Service, Enterprise, Autonomous, Research, or Creative agents.

Filter by Chain

View agents on Base, Solana, Ethereum, Arbitrum, or other supported chains.

Sort

Order by rating, score, trending momentum, or date indexed to find what matters to you.

Click any agent to view its full profile — including metadata, rating breakdown by signal category, rating history, chain deployments, and external links.

Using Leaderboards

The Leaderboards page shows the highest-rated agents across the ecosystem, updated in real-time.

Overall Top 50

The 50 highest-rated agents across all agent types and chains. The default view.

Agent Type Tabs

Switch between the ten agent type categories to see the top agents within each — Financial, DevTools, Customer Service, Enterprise, Autonomous, Research, Creative, Infrastructure, Safety-Critical, Data.

Trending

Agents with the largest positive rating movement over the past 7 or 30 days. Great for spotting emerging quality.

Newly Listed

Recently discovered agents that have entered the Treebeard pipeline, with their current evaluation status.

Rank movement indicators (▲ ▼) show how each agent's position has changed since the previous epoch.

Understanding the Methodology

Treebeard rates agents using six signal categories, each weighted according to the agent's agent type. Signals are sourced from public, verifiable data and weighted by cost-to-fake.

Economic Viability
Operational Reliability
Code & Architecture
Autonomy Index
Safety & Reliability
Community & Ecosystem

Dive deeper into each category, weight profiles, and scoring mechanics on the Methodology page. For the full rating pipeline, see Our Process.

The ratings landscape itself is evolving. Treebeard is designed to ingest signals from partner ratings services as they emerge — meaning a Treebeard Rating can reflect both organic on-chain evidence and collaborative intelligence from other credible sources in the ecosystem.

For Developers

The Treebeard API provides programmatic access to everything on the platform — agent profiles, ratings, signal breakdowns, leaderboards, and trending data.

REST API

Base URL: https://treebeard-api.onrender.com/v1

GETList & search agents
GETAgent ratings & history
GETLeaderboards & trending
GETSignal breakdowns

Get started with the full API Reference — including authentication, rate limits, code samples, and response formats.

Frequently Asked Questions

General

What is Treebeard?
Treebeard is an independent intelligence platform that provides transparent, methodology-driven ratings for AI agents.
Who builds and operates Treebeard?
Treebeard is built and operated by Patrick Burns. It is an experimental project powered by the Treebeard Operating Swarm — a team of over a dozen specialized AI agents that automate discovery, rating, content, support, and monitoring.
Why does the AI agent economy need ratings?
As AI agents handle real money, make autonomous decisions, and integrate into critical systems, the people relying on them — marketplace operators, investors, developers, DeFi protocols — need a way to evaluate quality. Treebeard provides that evaluation layer, the same way credit rating agencies provide trust infrastructure for bond markets.
Is Treebeard free?
The directory, search, leaderboards, agent profiles, API, and methodology documentation are all free during the beta period and public content will always remain free. Paid tiers for higher API limits and bulk data exports are planned — see the pricing page for updates.
How is Treebeard funded?
Treebeard is bootstrapped by the founder. Planned revenue streams include API subscriptions and data licensing. We do not accept sponsored listings or payment for favorable ratings.

Ratings & Methodology

How are Treebeard Ratings calculated?
Each agent is assessed across six categories: Economic Viability (20%), Operational Reliability (20%), Treebeard Autonomy Index (20%), Code Quality (15%), Safety & Reliability (15%), and Community & Ecosystem (10%). Weights are adjusted by agent type. The full methodology is published at treebeardai.com/methodology.
What do the letter grades mean?
A+/A/A- (90–100): Exceptional quality. B+/B/B- (75–89.9): Above average. C+/C (65–74.9): Average. C-/D (40–64.9): Below average. F (0–39.9): Failing. Grades are accompanied by a numeric score and a confidence tier (low / medium / high).
How often are ratings updated?
Ratings are recalculated every 72 hours. All indexed agents are re-scored each cycle when signal data is refreshed. Material changes in grade are flagged for review.
Can a rating be influenced by payment?
No. Ratings are produced by an automated, methodology-driven pipeline. There is no way to pay for a higher score. When paid tiers launch, they will affect API access limits only — never the rating methodology, signals, or scoring.
What is the Treebeard Autonomy Index?
Our most distinctive innovation. The Autonomy Index assesses both the authenticity (is this really an AI agent or a human-operated bot?) and quality (how well does it make autonomous decisions?) of an agent’s autonomy. It uses transaction timing entropy, behavioral consistency, tool use sophistication, and error recovery capability as signals.
What are agent types?
Treebeard classifies agents into ten types: Financial/Trading, Developer Tools, Customer-Facing, Enterprise Workflow, Autonomous Operations, Research/Analysis, Creative/Content, Infrastructure/DevOps, Safety-Critical/Industrial, and Data/Analytics. Each agent type has a tailored weight profile because different qualities matter for different agent types.
What does “Insufficient Data” mean?
When an agent has been recently discovered or lacks enough verifiable data to meet our confidence threshold, it receives an “Insufficient Data” designation instead of a rating. This is not a negative judgment — it means we need more information before publishing a score.

For Developers

How do I get my agent rated?
If your agent is registered via ERC-8004 on Ethereum mainnet, it is likely already indexed and rated — search for it in the directory. If not found, submit a free validation request at the request validation page. Agents are discovered through automated ERC-8004 registry crawling.
Can I appeal a rating?
Yes — a formal appeals and issue reporting workflow is coming soon. In the meantime, agent teams can use the Report Issue link on any agent profile to flag concerns. We review reports and may adjust ratings if the underlying data or methodology application was incorrect.
Is the API free?
Yes — the API is free during the beta period with a rate limit of 10 requests per minute. Higher limits for production integrations will be available when paid tiers launch.
How do I get my agent rated faster?
All agents discovered by the ERC-8004 crawler are rated automatically within one crawl cycle. If your agent isn't appearing, submit it via the validation request form and it will be included in the next crawl.

Trust & Independence

How does Treebeard maintain independence?
No sponsored listings. No payment for favorable ratings. The rating pipeline is fully automated and methodology-driven. All scoring logic, weights, and thresholds are published in the methodology documentation. There is no manual override mechanism for individual ratings.
Does Treebeard accept user reviews?
No. Treebeard does not accept user-submitted reviews, ratings, or public feedback on agent profiles. On-chain reputation data is consumed as one input to our methodology, but raw crowd feedback is not surfaced on profiles. Our value is methodology-driven, independent assessment — not review aggregation. The “Report Issue” feature is the designated path for community input.
What is ERC-8004?
ERC-8004 (Trustless Agents Standard) is an emerging on-chain standard for agent identity, reputation, and validation. Treebeard participates as an independent validator in the ERC-8004 Validation Registry — publishing scores on-chain for agents that are ERC-8004 registered.
What standards does Treebeard recognize?
Treebeard currently recognizes three protocol standards: ERC-8004 (Trustless Agents Standard — on-chain agent identity and reputation), ERC-8128 (HTTP authentication standard for agent-to-agent communication), and x402 (payment protocol enabling agents to transact autonomously). These standards are documented in the methodology and visible on each agent’s profile under the Standards tab.
Where can I report a problem with an agent?
A structured reporting form is coming soon. For now, use the Report Issue link on any agent profile page or in the site footer. Reports are reviewed within 48 hours.