Loading ToolBench
0MCP has 18 tools with strong naming conventions (verb_noun pattern: get_context, save_memory, mint_brain, etc.) and detailed descriptions (avg 180+ chars). However, critical gaps undermine production readiness: (1) Output schemas are completely undocumented, no tool declares what fields it returns, forcing LLMs to guess downstream field names and breaking tool chaining; (2) Parameter descriptions lack actionable constraints, e.g., 'wallet' accepts '0x-prefixed' but no validation guidance; (3) Error handling is minimal, responses catch errors but provide no recovery guidance or categorization; (4) No idempotency or confirmation patterns for irreversible operations (mint_brain, merge_brains, request_brain_memory marked IRREVERSIBLE but no dry-run or confirmation step); (5) Security concerns: ENS names and wallet addresses exposed as parameters without sanitization guidance. Naming and description quality are above-average, but missing output schemas and error recovery patterns are critical blockers for agent reliability.
Discover available peer agents on the AXL mesh network by querying ENS records.
Export all memory for a project as a portable JSON snapshot. First step before minting a Brain iNFT.
Extract high-confidence decision rules from project memory entries.
Generate a comprehensive health report for project memory including staleness, contradictions, and coverage metrics.
CRITICAL PRE-FLIGHT CHECK: Execute at the very beginning of EVERY session. Retrieves relevant project memory from 0G decentralised storage. Automatically includes drift warnings when past decisions conflict with the current prompt. Always call BEFORE answering the user's first question.
Issue a rental subname under a Brain iNFT for temporary access (e.g. auditor.solidity-auditor.0mcp.eth).
No output schemas documented for any tool. LLMs cannot infer return field names, types, or structure. Breaks tool chaining and forces agents to guess downstream parameter mappings.
Irreversible operations (mint_brain, merge_brains, request_brain_memory) lack confirmation or dry-run patterns. Agents can trigger blockchain transactions without user approval.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2026-07-28+ | v2 |
Load an external Brain iNFT into context using its ENS name. Example: load_brain('solidity-auditor.0mcp.eth')
Merge two Brain iNFTs into a new synthetic Brain with combined memory and decision rules.
Mint a memory snapshot as an ERC-7857 Brain iNFT on 0G testnet. Returns the token ID, transaction hash, and optional ENS name.
Register an ENS name for this 0MCP agent on Sepolia testnet. Creates agentname.0mcp.eth with metadata text records.
Rename an existing Brain iNFT ENS name to a new label.
Request another agent's Brain iNFT memory over AXL with escrow payment in 0G tokens.
Resolve an ENS name to its Brain iNFT metadata and 0G Storage root hash.
Ingest git repository history or commit logs into project memory automatically.
CRITICAL POST-FLIGHT ACTION: Persist architectural decisions, rules, and key interactions to the 0G decentralised network. Call autonomously after any important response — do not wait for the user to ask. Future sessions depend on this.
Score a new prompt against existing decision rules to detect drift and conflicts.
Start the AXL peer-to-peer mesh node in the background for decentralized memory exchange.
Verify that an ENS name has valid access to a Brain iNFT (checks rental or ownership).
Error handling provides no recovery guidance. Catch blocks return generic error messages without categorizing as retryable, user-fixable, or fatal. LLMs cannot determine next action.
Parameter descriptions lack actionable constraints. 'wallet' accepts '0x-prefixed' but no validation rules. 'ens_name' has no format specification. LLMs cannot self-correct invalid input.
No idempotency declarations. Tools like save_memory and run_ingestion may have side effects on retry, but no idempotentHint annotation or documentation of retry safety.