A .NET-based MCP server providing memory management, bridge relay, and orchestration tools for multi-agent fleet systems
Phleet defines 9 tools with consistent naming (verb_noun pattern) and reasonable descriptions (avg ~120 chars). All tools have input schemas with type definitions and parameter descriptions. However, output schemas are not documented in the source, parameter constraints lack enums/ranges, and error handling guidance is absent. Memory tools are well-composed (search, list, get, store, update, delete) but lack pagination specs. The ask_acto tool accepts a timeout but provides no guidance on retryability or failure modes. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications in the spec.
Send a question to Acto (fleet CTO) and wait for a synchronous response. Use this to get architectural guidance, cross-project context, or fleet knowledge.
Check if Acto (fleet CTO) is idle or busy. Use before sending a long question to avoid waiting on a busy agent.
Delete a memory. By default, moves it to _archived/ (soft delete). Use permanent=true to permanently remove it.
Get the full content of a specific memory by its ID. Use this after memory_search or memory_list to read the complete memory.
List memories with optional filters. Returns a summary of matching memories without full content. Use memory_get to read full content.
Semantic search across all memories. Returns the most relevant memories matching the query. Use this to find prior knowledge, past decisions, error resolutions, and learnings.
Output schemas not documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., memory IDs from search results for chaining to memory_get). Every tool should declare return type structure.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk classifications. memory_delete is DESTRUCTIVE, memory_store/memory_update are WRITE, these should be marked in the schema to guide agent planning.
Pagination not specified for list/search tools. memory_list and memory_search accept 'limit' but no offset/cursor or total_count guidance. Large result sets will blow context windows without pagination.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 65 | 2026-07-28+ | v2 |
Get statistics about stored memories — counts by type, total count, and archived count.
Store a new memory. Creates a markdown file and indexes it for semantic search. Use this to persist learnings, task results, decisions, error resolutions, and other knowledge.
Update an existing memory's title, content, tags, or project. The memory will be automatically re-indexed for search.
No error recovery guidance. ask_acto accepts timeout_seconds but provides no documentation on what happens on timeout, whether to retry, or how to handle Acto unavailability. memory_delete's 'permanent' flag defaults to false but lacks confirmation/dry-run pattern for destructive operations.
Parameter constraints missing. memory_store's 'type' enum is documented in description but not enforced as JSON Schema enum. memory_list's 'limit' has no min/max bounds. Agents cannot validate inputs before calling.