Algorand Model Context Protocol Local Implementation for blockchain transaction management, account operations, and ecosystem integrations
The Algorand MCP server defines 9 tools with schemas and descriptions, but exhibits inconsistent quality across the board. Naming follows verb_noun convention (api_nfd_*, api_algod_*) but lacks clarity in distinguishing similar operations. Descriptions are present for all tools (100-200 chars range, within baseline), but parameter descriptions are sparse or missing. Input schemas are well-formed JSON Schema with type definitions and enums where appropriate. However, output schemas are not documented, the server returns structured data but LLMs cannot plan downstream calls without knowing what fields to expect. All tools are read-only (appropriate for a blockchain query interface), but error handling is minimal and does not guide recovery or provide actionable next steps. The resource feature is declared but not deeply integrated with tool composition.
Get application box by name
Get all application boxes
Get application information
Browse NFDs with various filters
Get a specific NFD by name or by its application ID
Get activity/changes for NFDs
Get analytics data for NFDs
Output schemas are not documented. Tools return structured data but LLMs cannot plan downstream calls without knowing response fields, cardinality, or types. Without documented return schemas, agents cannot compose these tools with confidence or extract chaining IDs for follow-up calls.
Parameter descriptions are inconsistent or missing detail. Many parameters lack guidance on valid ranges, formats, or when they should be used. For example, 'view' enum has options (tiny, brief, full) but no description of what each returns or when to use which. 'limit' and 'offset' appear repeatedly without min/max bounds or guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 61 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 32 | - | v1 |
Get NFDs for specific addresses
Search NFDs with various filters
No error handling guidance. Tool definitions do not specify what errors are possible, which are retryable, or what the agent should do next. No recovery paths for common failures (e.g., NFD not found, invalid application ID).
No pagination guidance in tool descriptions. Several tools accept limit and offset, but descriptions do not explain max limits, whether results are ordered, or what happens if offset exceeds available results. RFC baseline: paginated results should document total count and next_cursor behavior.
Parameter naming inconsistency across similar tools. Some tools use 'name' (string), others use 'address' (array). Mixing singular and array types for semantically similar operations (e.g., api_nfd_get_nfd takes nameOrID as string; api_nfd_get_nfds_for_addresses takes address as array). LLMs may conflate these and pass wrong types.
No tool annotations (readOnlyHint, idempotentHint). Per MCP spec 2026-07-28, tools should declare whether they modify state, require confirmation, or are safe to call repeatedly. All tools here are read-only but this is not machine-declared, forcing LLMs to infer from naming alone.
Similar tool names with unclear distinction. 'api_nfd_browse_nfds' and 'api_nfd_search_nfds' sound nearly identical, descriptions do not clearly explain when to use browse vs search. Are they the same query with different filters, or fundamentally different operations?