A Model Context Protocol (MCP) server for interacting with the Tally API, enabling AI agents to access DAO governance data
This server presents a moderately strong foundation with clear, action-oriented tool names and well-structured parameter schemas. All 10 tools are explicitly defined with input schemas, enums, and type constraints. Descriptions are present and functional but tend toward brevity (averaging ~100 chars); many lack context about WHEN to use the tool or multi-step workflows. Parameter descriptions are generally adequate but inconsistent, some are detailed ('The DAO's slug (e.g., 'uniswap' or 'aave')') while others are minimal ('The address to get votes for'). No output schemas are documented, which forces LLMs to infer response structure. Error handling and recovery guidance are absent from the visible source. The server is read-only (all tools are queries), reducing risk but also eliminating the need for idempotency or confirmation patterns. Naming follows the verb_noun pattern well (list-*, get-*, etc.), but some parameter names lack type suffixes (e.g., 'address' should hint it's Ethereum format, which the description partially does).
Get proposals created by an address for a specific organization
Returns proposals from DAOs where a given address has participated (voted, proposed, etc.)
Returns delegations received by an address
Get votes cast by an address for a specific organization
Get detailed information about a specific DAO
Get list of delegators for a specific address
No output schemas documented. LLMs cannot infer what fields are returned by each tool, forcing them to guess response structure and blocking downstream tool chaining.
Tool descriptions are too brief (avg 100 chars, target 150-200). Most descriptions answer 'what does it do' but not 'when to use it' or 'how does it differ from similar tools'. LLMs struggle to select the right tool among get-address-votes, get-address-created-proposals, and get-address-daos-proposals without better disambiguation.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 60 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 39 | - | v1 |
Get detailed information about a specific proposal. You must provide either the Tally ID (globally unique) or both onchainId and governorId (unique within a governor).
List DAOs on Tally sorted by specified criteria
List delegates for a specific organization with their metadata
List proposals for a specific DAO or organization using its slug
Parameter descriptions lack context about input format and constraints. Example: 'address' is described as 'The address to get votes for' but should specify '0x-prefixed Ethereum address (40 hex chars)' for clarity. Similar issues with 'organizationSlug' and 'slug', no detail on valid format or resolution.
Pagination support is present (limit, afterCursor, beforeCursor) but not consistently implemented. list-daos and list-proposals support both afterCursor and beforeCursor, but get-delegators supports all three; others omit beforeCursor. No indication of what constitutes a 'cursor' format (opaque string? ID? timestamp?).
get-proposal uses 'oneOf' with two alternative schemas (id vs onchainId+governorId). This is valid JSON Schema but most LLM tool callers do not understand conditional constraints. LLMs may pass all parameters or none, violating the oneOf contract. Should document the constraint explicitly in the description: 'Provide either id OR (onchainId AND governorId), not both.'
No error handling or recovery guidance visible in tool definitions. If a DAO slug is invalid, or an address has no votes, what does the tool return? An error message? An empty list? A 404? LLMs need guidance on what to do next.