Nehemvix vs DeepWiki: what each is for
DeepWiki gives free wikis and answers for public GitHub repositories, with no model key of your own and a free MCP server that needs no login. Nehemvix writes a checked guide of any Git repository, public or private, with your own model key, and shows the cost before it spends anything. This comparison is as of 5 November 2026.
Two tools for the same first question
Both tools answer the question every developer has in front of an unfamiliar repository: what is this, how is it put together, and where do I start? They answer it in different ways, for different situations, and many people will find a use for both.
This page states what each one does from its maker's own pages, links to the source of each claim about DeepWiki, and says nothing about DeepWiki that its pages do not say. Products change quickly; the facts below were checked on 5 November 2026. If something here is out of date, tell us and we will correct it.
What DeepWiki is
DeepWiki is made by Cognition, the company behind the AI software engineer Devin. According to Devin's DeepWiki documentation:
- it indexes repositories and produces wikis with architecture diagrams, links to the source and summaries of the codebase;
- a free version for public GitHub repositories runs at deepwiki.com, with wikis and questions answered by Ask Devin, and anyone can submit a public GitHub repository for indexing;
- inside Devin, wikis are generated when repositories are connected, a repository can steer which pages are written with a
.devin/wiki.jsonfile, and the default effort level is free while higher levels are billed against Devin credits.
DeepWiki also has an MCP server. Cognition's announcement (22 May 2025) describes it as free, with no login or authentication, and lists three tools: read_wiki_structure, read_wiki_contents and ask_question. The MCP documentation gives the endpoint as https://mcp.deepwiki.com/mcp, says it serves public repositories, and points to the separate Devin MCP server, which needs a Devin account and API key, for private ones.
The strengths are plain: for a public GitHub project there is nothing to pay and no model key to bring, and an assistant can reach it over MCP without a login.
What Nehemvix is
Nehemvix is made by AELSHA™. It turns a Git repository into a documentation guide for a chosen audience (a beginner, a senior engineer or an architect), with architecture and data-flow diagrams and citations that link to exact lines at the analysed commit.
- Any HTTPS Git host: GitHub, GitLab, Bitbucket or your own server. Private repositories work with a read-only access token that is used once and never stored.
- Your own model key: Nehemvix is free, and you pay your AI provider directly. Before anything is spent it shows a low, likely and high cost, and you approve it. On 3 November 2026 a guide of a 46-file Python library for all three audiences was estimated at $0.036 to $0.077 and cost $0.0459.
- Checked output: every file and function name in the guide is looked up in the code, citations are resolved by code rather than written by the model, and each guide gets a grounding score from 0 to 100 with anything it could not verify listed. The checker itself is measured on 12 reference guides, whose facts a person checked by hand, and on blind lists of invented names; the numbers are on how we check.
- Exports: Markdown, HTML, PDF, Word, PowerPoint and editable draw.io diagrams, from the web app or the
nehemvixcommand line. - AI assistants: an MCP server and an A2A endpoint with seven read-only tools over your own guides. They read and search; there is no model on Nehemvix's server, so the assistant you use does the answering, and an assistant cannot start a paid guide.
Side by side
| DeepWiki | Nehemvix | |
|---|---|---|
| Free to use | Yes, for public GitHub repositories | Yes; you pay your own model provider |
| Model key needed | No | Yes, from a provider you choose |
| Repositories | Public GitHub at deepwiki.com; private through Devin | Any HTTPS Git host, public or private |
| What you get | A wiki per repository, and questions answered | A guide per audience, with a grounding score |
| Assistants over MCP | Three tools, including AI answers | Seven read-only tools, no model on the server |
The table holds only what each maker's pages state. Where a cell is short, read the sections above for the detail.
Which one fits
Choose DeepWiki when you want to look around a well-known public GitHub project at no cost and without a model key, or you want your coding assistant to ask questions about a public library through a free MCP server. If your team already works in Devin, its wikis come with the repositories you connect.
Choose Nehemvix when:
- the code is private, or lives on GitLab, Bitbucket or your own Git server;
- you need to decide which model provider sees the code, and want consent recorded for each one before anything is sent;
- you want a document for a specific reader, such as onboarding notes for a new hire or an architecture review, rather than a general wiki;
- you need to know how far to trust it: a score, the list of names it could not verify, and citations you can click;
- you want files to hand on: a PDF, a Word document, slides or an editable diagram.
Some people will use both: DeepWiki to browse an open-source dependency, and Nehemvix for their own code. Neither replaces reading the code. What each tells you is a map, and how to understand a new codebase shows how to check a map against the territory.
How we kept this fair
We are the makers of Nehemvix, so this page holds to a few rules. Every claim about DeepWiki links to Cognition's own pages. We do not guess at features those pages do not mention, and we do not compare quality, because we have not measured DeepWiki's output with the same evals we use on our own. Claims about Nehemvix come from its documentation and its published eval record.
How Nehemvix writes and checks a guide is on how it works, and the AI assistant setup is in connect an AI assistant. For why a diagram drawn from the code beats one from memory, see draw an architecture diagram from your code.
Learning this to build your own agent? AELSHA™ can help you design it.