@kleros/agentkit is a command-line tool and MCP server for reading Kleros data. Both AI agents and humans can use it. One interface covers courts, disputes, evidence, court and dispute policies, and juror stakes on Kleros V1 and V2.
Current release: v0.3.0 (September 2026). The CLI is read-only and pre-1.0, so its commands may change between minor versions. Write operations (staking, voting, creating disputes, submitting evidence, appeals) are not available yet.
What it can do
- Browse courts, disputes and juror activity
- Resolve court and dispute policy: the rules jurors are asked to apply
- Walk evidence submissions, including attachments hosted on IPFS
- Classify arbitrable contracts and pull a combined dispute brief in one call
- Rank jurors by staked PNK
cta block that suggests the next logical command.
Supported chains
A list command run without
--chain fans out across Ethereum, Gnosis and then Arbitrum One. V1 and V2 do not mean the same thing by ruled, so each dispute list row carries a ruledSemantics marker. Use dispute get when you need a ruling read live from the arbitrator contract.
Quickstart
Requires Node.js 22 or later. Run it without installing:kleros command:
--chain every time:
Use it as an MCP server
Register the CLI as an MCP tool server so agents in Claude Code, Cursor and other MCP clients can call dispute queries directly:kleros mcp add registers it for Claude Code and Cursor. Tool names follow the group_command pattern, for example dispute_list, court_get and juror_info. Pass a chain argument on any tool call to choose the chain for that call.
For agent discovery, kleros --llms prints a compact command manifest and kleros --llms-full prints full arguments and output schemas. kleros skills add installs the bundled agent skill.
Production use
The default RPC endpoints are free public ones and may rate-limit under real traffic. For reliable use, point theKLEROS_RPC_URL_<CHAIN> overrides (for example KLEROS_RPC_URL_ARBITRUM_ONE) at a provider you trust. On Ethereum, use an archive-capable endpoint, because resolving older policies needs historical logs.
Resources
npm package
@kleros/agentkit releases and READMEKleros Skills
Agent-readable knowledge packs for Kleros
Report an issue
Feedback goes to the public kleros-skills repository
Subgraph endpoints
The data sources the CLI reads from