Four tools with complete input schemas and descriptions, but descriptions are minimal (10-20 chars baseline) and lack LLM-optimized guidance. All tools are READ_ONLY and operate on a local project state. Naming follows verb_noun pattern (morpheus_get_*, morpheus_check_*). No output schemas documented. Parameters are well-typed (string) but lack format constraints or validation guidance. No error handling patterns visible. Tools appear functional but lack production-grade polish for agent reasoning.

• Expand tool descriptions to 100-200 chars. For morpheus_check_text: 'Verify claims in input text against the active Morpheus state without making external provider calls. Returns a list of verified/unverified claims with supporting evidence. Use this to validate agent-generated text before committing it.'
• Document output schemas for each tool. E.g., morpheus_get_active_state returns {claims: [{text, verified, evidence_ids}], sources: [{id, url, title}], evidence: [{id, claim_id, source_id, excerpt}]}.
• Add format constraints to project_root parameter: 'Absolute or relative path to a Python project with a pyproject.toml file. Must be readable by the server process.'
• Implement error handling with recovery hints. E.g., if project_root does not exist, return 'Project root not found at {path}. Verify the path is correct and the directory is readable.'
• Add a discovery tool (morpheus_list_projects or morpheus_describe_state_schema) to help agents understand available state structure before calling specific tools.