In-depth architectural comparison of the Agent Harnesses and Greenlit Books MCP servers. Compare execution transports, security boundaries, tool capabilities, quality scores, and ready-to-paste client installation snippets for Claude, Cursor, Windsurf, and VS Code.
At a Glance & Executive Verdict
Agent Harnesses
Developer Tools · Local stdio
Quality: 48/100 (Fair) | Auth: No auth required
Greenlit Books
Developer Tools · Remote HTTP/SSE
Quality: 52/100 (Good) | Auth: No auth required
Verdict Summary: Choose Agent Harnesses if you need specialized Developer Tools tools running via a local process. Choose Greenlit Books if your workspace requires Developer Tools integration with remote web transport. Both servers can be configured concurrently in your client's mcpServers manifest.
Which MCP Server Should You Choose?
Choose Agent Harnesses when:
You need dedicated capabilities in the Developer Tools domain.
You prefer local stdio subprocess transport architecture.
Your security boundary fits: No auth required (Free / Open Source).
Ask a question in your own words and get the passages that answer it, from what the books publish free: chapter one of every book (by section), each book's named concept and its FAQ, the glossary, the claim ledger and the guides. Matched by words and by meaning. Each passage comes with the exact UR…
Ready-to-Paste Client Configurations
Paste either (or both) of these JSON server blocks into your client config file (e.g. claude_desktop_config.json or ~/.cursor/mcp.json).
Agent Harnesses is categorized under Developer Tools and uses a local stdio subprocess. In contrast, Greenlit Books belongs to Developer Tools using remote streaming HTTP/SSE transport. Select Agent Harnesses when you need capabilities focused on developer tools and Greenlit Books when you require tools for developer tools.
Search everything on greenlitbooks.com: safety verdicts on AI agents, MCP servers and coding tools ("is X safe"), field notes, guides and the books. Returns ids to pass to `fetch`.
fetch
Read one page by an id from `search`: the page as markdown, with its canonical URL to cite and its dates.
search_greenlit_books
Ranked full-text search over the catalog. Filters: `series`, `audience`, `limit`.
get_book
One book in full: chapters, word count, free chapter one, Amazon Kindle and paperback links, prices, Kindle Unlimited status, related reading.
find_books_for_topic
Match a topic phrase (agent reliability, prompt injection, human in the loop, durable execution, ...) against the curated topic map.
get_reading_path
One series with its books in reading order.
get_concept
The named question a book answers, with its short answer, full explanation, and a links block (free chapter, Amazon, cite-as).
get_free_chapter
The complete text of chapter one of any book, free to read and to quote, with the chapter list and where to get the rest.
cite_source
The finished citation for a book, one of its free chapters, or one of its coined terms: the line to print, the canonical URL, the markdown twin of the same page, and the date it was last verified. It asserts nothing about the content, only how to name the source.
define_term
The glossary: a term the books coin or pin down (the green lie, blast radius, gate faith), defined in the book's own words, with the chapter, the other books that use it, a checkable claim, related terms, and links to cite and hand off. Unknown term returns the full list with aliases.
check_claim
Search the claim ledger for claims bearing on a statement, before asserting it. Each hit carries the claim, a citation URL resolving to that exact sentence, what kind of statement it is and the one-line rule for quoting that kind, a citation already written in prose and BibTeX, and for every extern…