Loading ToolBench
MCP server for Stellar Intel — provides tools for off-ramp rate comparison, intent preparation/execution, anchor reputation, health monitoring, and corridor analysis across Stellar anchors
Stellar Intel MCP has 8 well-named tools with clear action verbs (quote, prepare, execute, probe-coverage, anchor-reputation, anchor-health, leaderboard, corridors). All tools have descriptions (avg ~90 chars, within baseline 34-392 range). Input schemas are present with typed parameters and descriptions. However, output schemas are not documented in the provided source, and parameter descriptions lack detail on constraints, formats, and valid ranges. Error handling guidance is minimal. The execute tool (WRITE risk) lacks confirmation/dry-run pattern. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classification.
Get health status and operational metrics for a Stellar anchor
Get reputation scores and settlement history for a Stellar anchor from the on-chain oracle
List available off-ramp corridors with supported anchors and asset pairs
Submit a signed off-ramp transaction to a Stellar anchor and track the intent
Get anchor leaderboard rankings by reputation score, settlement volume, or success rate
Prepare an off-ramp intent for execution by validating anchor capabilities and building transaction parameters
Check probe coverage and readiness status for corridors and anchors
Get off-ramp rate comparison across Stellar anchors for a given corridor and amount
Output schemas not documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., does quote return anchorId for prepare chaining?). Baseline: 100% of A+ tools document return types.
Parameter descriptions lack constraint details. 'amount' (quote, prepare) does not specify format (decimal places, min/max, currency). 'sortBy' (leaderboard) lacks enum constraint. 'country' (corridors) lacks format hint (ISO 3166-1 alpha-2?). Baseline: 94% of A+ tools include format/range in param descriptions.
execute tool (WRITE risk) lacks confirmation/dry-run pattern. Agents can submit signed transactions without a preview step, risking irreversible errors. No error recovery guidance (e.g., 'If submission fails, call probe-coverage to verify anchor status').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Risk classification exists in metadata but not exposed to MCP protocol. Agents cannot distinguish safe vs. destructive tools without reading descriptions.
No pagination support documented. leaderboard accepts 'limit' but no offset/cursor or total_count return. Large result sets risk context window exhaustion. Baseline: tools returning lists should support pagination.