What is the VerseDB MCP?

The VerseDB MCP lets you drive your VerseDB account from Claude Desktop, ChatGPT, Claude Code, the Gemini API, or any other Model Context Protocol client. Ask your AI assistant to search the database, update your collection, manage your pull list, create lists, or look up market prices — it calls VerseDB on your behalf. Pro-only, gated by OAuth or a personal API token.

Setting up your AI assistant#

The server lives at https://versedb.com/mcp/api. Interactive clients (Claude, ChatGPT, Claude Code) handle authentication for you via OAuth on first connection and refresh automatically. SDK-based clients (the Gemini API and other code-driven setups) authenticate with a personal API token you create from your account. Rate-limited to 1,000 requests per hour per user.

Claude Desktop. Settings → Connectors → Add custom connector. Give it a name like "VerseDB" and the URL above. Claude opens your browser to sign in, you approve the app, and the connector is ready.

ChatGPT. Settings → Connectors → Add custom connector (requires Plus, Pro, Team, or Enterprise; turn on Developer mode if the option is hidden). Name it "VerseDB" and use the URL above. ChatGPT runs the OAuth flow in your browser the first time you ask it to do something with VerseDB.

Claude Code. Run:

claude mcp add --transport http versedb https://versedb.com/mcp/api

Claude Code walks you through OAuth the first time you use a tool.

Gemini API (Google Gen AI SDK). Gemini understands remote MCP servers natively, but SDK calls send a static Authorization header — there's no interactive OAuth handshake. Create a personal API token from your VerseDB account at Settings → Apps → Custom, then pass it through:

from google import genai

client = genai.Client()
response = client.models.generate_content(
    model="gemini-2.5-pro",
    contents="What's coming out on my pull list next week?",
    config={
        "tools": [{
            "mcp_server": {
                "url": "https://versedb.com/mcp/api",
                "headers": {"Authorization": "Bearer YOUR_VERSEDB_API_TOKEN"}
            }
        }]
    },
)

The JavaScript SDK (@google/genai) accepts the same mcp_server shape. Treat the token like a password — anything that calls the MCP with it can read and write your library.

Other clients. Anything that speaks the Model Context Protocol (Cline, Zed, Witsy, custom SDKs) works the same way. Point it at the URL, complete OAuth in the browser if the client supports it, or pass a personal API token in the Authorization header.

What you can do with the MCP#

Once your client is connected you have full Pro-level access — search, browse, and manage your library.

Search and browse. Find series, issues, creators, characters, publishers, teams, story arcs, events, tier lists, comic shops, and podcasts. Pull full details for any of those entities — the assistant starts with a broad search and then fetches what matters.

Your collection. List what you own, add issues with grade and condition, update items (re-grade, move to a box, add purchase price), and remove them. Works with any issue the assistant can find.

Reading progress. Mark single issues as read or unread, or check how far through a series you are.

Pull list. See what's on your pull list, add series to it, or remove them. Ask for "what's coming out next week" and the assistant combines your pull list with upcoming releases.

Lists. Create, rename, and delete lists. Add issues, series, characters, creators, story arcs, or teams. Share curated reading orders or themed collections.

Reviews. Write a review for an issue, update it, or pull your existing ones. One review per issue per account, half-star increments.

Market prices. Fair market values, recent sale comps, and price trends by grade for any issue.

Shops, events, and podcasts. Find local comic shops, upcoming conventions and store events, and comic podcasts across the site.

Tips and limitations#

Tell it who you are up front. The assistant can't see your VerseDB profile. Mention what you collect, your favorite publishers, or which runs you're working on — the tools produce much better answers with a little context.

Trust the IDs it returns. Every tool response includes the VerseDB ID for the entity. If you're making multiple changes to the same issue or series, referencing the ID instead of searching again is faster and avoids picking the wrong record.

One review per issue. The assistant can create a review only if you haven't written one for that issue. To change an existing rating, ask it to update the review.

It can't see your lists page or UI state. All writes go through tools, not the browser session. The assistant doesn't know what's currently on your screen.

It doesn't edit series, characters, or creator data. Contribution edits still go through the VerseDB moderation queue from the web UI. The MCP is read + personal-library only.

Rate limit. 1,000 requests per hour. Long back-and-forth research chats stay well under that, but a runaway loop can trip it.