An MCP server for interacting with WriteFreely API, enabling AI agents like Claude to perform blogging actions. Supports both self-hosted WriteFreely instances and the Write.as hosted service.
The WriteFreely MCP server presents a mixed quality profile. Strengths: all 10 tools have explicit names, descriptions, and documented input schemas with types. Tool naming follows verb_noun conventions (publish_post, edit_post, delete_post, etc.). Weaknesses: descriptions are inconsistent in depth, some tools (e.g., hello_writefreely, login) have good context, but others (e.g., browse_collection) lack guidance on when/why to use them. Parameter descriptions are present but terse; many lack examples of valid formats or constraints. No documented output schemas beyond scattered mentions in descriptions. Error handling is minimal, no guidance for recovery or categorization. The token/credentials design is problematic: tools accept access_token as a parameter, which violates secret-injection patterns and could leak credentials in logs. Most tools lack idempotency guarantees and destructive operations (delete_post) have no confirmation mechanism. Per-tool analysis reveals schema completeness ranges from good (publish_post) to minimal (hello_writefreely). Overall alignment with the 54 patterns is partial: composition is reasonable (single-concern tools), but security, error handling, and schema documentation lag.
List posts in a specific collection/blog.
Read posts from the public read.write.as feed.
Permanently delete a post.
Update an existing post's content and/or title.
Simple test tool - confirms the WriteFreely MCP server is running.
Get all blogs/collections owned by the authenticated user.
Retrieve all posts for the authenticated user.
Credentials exposed as tool parameters. login() and all authenticated tools accept access_token as a string parameter, violating secret-injection pattern. Credentials will appear in agent logs, traces, and prompt history.
No documented output schemas. Tools return responses via PostResponse, AuthResponse, etc. (visible in api_client.py), but these schemas are not documented in tool descriptions. LLMs cannot plan downstream calls without knowing response structure.
Destructive operations lack confirmation. delete_post() has no dry-run or confirmation step. Agents can permanently delete posts on first call without safety gates.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 53 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 44 | - | v1 |
Authenticate with WriteFreely and obtain an access token. If WRITEFREELY_ACCESS_TOKEN environment variable is set, this tool will use that token automatically without requiring username and password. You can provide username and password to authenticate with a different account, or use the environment variable for better security (recommended for production).
Create and publish a new post on WriteFreely. Supports anonymous posts or publishing to a specific collection/blog.
Retrieve the content of an existing post by its ID.
Minimal error guidance. Error responses (visible in api_client.py authenticate() return empty string on failure) do not tell LLMs what to do next, no recovery hints, categorization, or suggestions for alternative tools.
Terse parameter descriptions. Many parameters lack format examples, constraints, or context. E.g., 'skip' in browse_public_feed just says 'Number of posts to skip', no range, no default behavior on invalid values.
No pagination metadata in descriptions. list_my_posts(), list_my_collections() have no description of pagination limits, total count return, or next_cursor behavior. Agents cannot determine if they've retrieved all results.
Token parameter redundancy. Tools offer both WRITEFREELY_ACCESS_TOKEN env var AND access_token parameter. Description says 'optional if token is set via env var', but this requires the LLM to reason about environment state, better to hide env var and auto-inject server-side.
Tool descriptions lack 'when to use' guidance. E.g., browse_collection vs list_my_collections, both list posts, but descriptions don't clarify the distinction or when to prefer one.