Direct answer
Give Your AI Tools
Your AI can talk about work all day and still not do any of it, until you connect it to an MCP server. Connecting Zambo at https://zambo.dev/mcp gives the AI you already use real tools, 153 in total with 35 on the default surface, plus memory that survives the session, continuity when you switch AIs mid job, and a verifiable receipt for every tool call. There are three ways to connect, an in-app connector, a config file, or one command, and this page explains each and routes you to the exact guide for your client.
Updated October 20, 2026.
What giving your AI tools actually means
A language model on its own can only talk. It can draft, summarize, reason, and explain, but it cannot check a live price, open a page as it exists today, run an audit, save a fact somewhere durable, or do any of the other verbs a working day is made of. Giving your AI tools means connecting it to an MCP server: a hosted service, reached at one URL, that offers named tools the model can call during a conversation or an agent session. MCP, the Model Context Protocol, is the common wiring for this, and it is why the same server can plug into ChatGPT, Claude, Cursor, VS Code, Windsurf, Codex, Claude Code, Gemini, Grok, Cline, and Zed without being rebuilt for each one.
The connection is small and the change is large. Before it, asking for a current figure gets you a confident answer from training memory. After it, the AI can look the figure up through a tool, show you the result, and file a receipt for the call. Before it, everything the AI learns about your project evaporates when the session ends. After it, facts you choose to save are held on your account and recalled in the next session, in the same AI or a different one. This page explains the four things a connection gives you, the three ways connections are made, and then routes you to the exact per client guide for the AI you actually use, because the last step is always client specific even though the idea is universal.
The four things a connection gives an AI
Tools. Real capabilities beyond conversation: web and research work, data and file jobs, site checks and audits, live prices, lead finding, and the rest of a working catalogue. Zambo runs 153 tools in total behind one endpoint, https://zambo.dev/mcp, with 35 on the default surface so a session loads the weekly set instead of the whole workshop.
Memory. Facts you explicitly save, with memory_remember, are stored on your Zambo account and return with memory_recall when you ask. Zambo memory holds only what you tell it to save; it does not import your chats or guess at what matters. A preference, a project constraint, a client detail: saved once, present everywhere you connect.
Continuity. Work state saved in one AI can be opened in another AI connected to the same account. Start a job in ChatGPT, hit a limit or a better idea, continue it in Claude or Cursor with the state intact. Continuity turns a row of separate apps into one workspace with several doors.
Receipts. Every saved tool result is filed with a verifiable receipt in the open AER-1 format, draft-zambo-aer1-24. Anyone can check a receipt free at https://zambo.dev/verify/ with no signup. A receipt proves the saved result was not changed after the fact; it never proves the answer was right, and that honest split is what makes the record worth trusting.
On top of the four sits the delivery shape they make possible: handover, one link carrying a finished job, its goal, steps, outputs, and receipts, that a client or teammate opens with no account and accepts or flags. Tools do the work, memory and continuity carry it between sessions and AIs, receipts make it checkable, and handover hands it over. See the shape live before connecting anything: the live demo runs one real call and hands you its receipt in about 10 seconds, and zambo.dev/today shows a recorded day of work.
Way one: the in-app connector
Chat based AIs connect from inside the app. In ChatGPT and Claude, you open the connectors or settings area, add a custom connector, and paste the server URL, https://zambo.dev/mcp, with no key to enter. Grok takes custom connectors the same way. The appeal is obvious: no files, no terminal, and the tools appear in the conversation where you already work. The per client details still matter, the exact menu names, where the connector list lives, and how you confirm the tools loaded, which is why the routing section below links a dedicated guide for each app instead of pretending one screenshot fits all of them. If your AI offers a connector gallery, a custom URL entry is the route to use for Zambo; the endpoint is the same one every other way on this page uses.
Way two: the config file
Editors and editor based agents connect through a small config file you edit once. Cursor, VS Code, Windsurf, Cline, and Zed all follow this pattern, and the differences are exactly the kind that waste an afternoon: the file has a different name and location in each client, and the address field is spelled differently, url in some, serverUrl in Windsurf, httpUrl in Gemini CLI settings, with VS Code also naming its top level key servers instead of mcpServers. The shape underneath never changes: one entry, one name, one address, https://zambo.dev/mcp, no local process because the server is hosted. Paste, save, reload or refresh, confirm the tools are listed, then run one checkable call. Each per client guide below gives you the exact file, the exact block, and the exact verification step for that client, verified against how that client actually reads its config.
Way three: one command
Terminal tools connect from the command line. Codex takes codex mcp add zambo --url https://zambo.dev/mcp, which writes the config entry for you. Claude Code and Gemini CLI have their own one line or settings routes to the same endpoint. The command route suits people who live in a terminal and teams who script their setups: the connection becomes a line in a setup script or a runbook instead of a file edited by hand on each machine. Whichever of the three ways your client uses, what lands on the other side is identical, the same endpoint, the same account, the same tools, memory, continuity, and receipts, which is why learning the pattern once carries across every AI you adopt later.
Pick your client: the exact guides
ChatGPT: connect through the in-app connector. Follow add an MCP server to ChatGPT for the exact steps.
Claude: connect through the in-app connector on web or desktop. Follow add an MCP server to Claude.
Cursor: connect through the MCP config file. Follow add an MCP server to Cursor.
VS Code: connect through mcp.json under the servers key. Follow add an MCP server to VS Code.
Windsurf: connect through mcp_config.json with the serverUrl field. Follow add an MCP server to Windsurf, or the Zambo specific add Zambo to Windsurf.
Codex: connect with one terminal command or the config.toml block. Follow add an MCP server to Codex.
Cline and Zed: connect through their settings files. Follow add Zambo to Cline or add Zambo to Zed.
Gemini CLI: connect through settings.json with the httpUrl field. Follow add Zambo to Gemini CLI.
What a good first session looks like
Whichever way you connected, judge the setup by one complete loop, not by a settings screen. Ask your AI to do something small through a tool, a live lookup or a page audit, and read the actual result instead of accepting its summary of the result. Ask it to save one fact that matters to your work, a preference, a project rule, a name and role, and then ask for it back, in the same session and, if you have a second AI connected, from that one too. Open the receipt link on the tool call and run the free check at zambo.dev/verify. Then do one piece of real work and, if it is for someone else, send it as a handover link rather than a paragraph claiming it is done. Tools, memory, receipts, delivery: one session, four proofs the connection is real.
Safety, in plain terms
Connecting a server gives the model tools it can call, so connect servers you can explain. Read any config block before pasting it, including blocks from this hub. Prefer hosted servers reached by one URL, like https://zambo.dev/mcp, over local servers that run unknown code on your machine, and be suspicious of any guide that asks for broad keys or credentials a plain connection does not need: Zambo needs no API key to connect. Keep the server list short, one broad server plus a specialist only where a real job demands it, and remove entries you no longer use, because every connected tool is attention the model spends and access you have granted. Memory deserves the same discipline: save facts deliberately, review what you saved, and remember that recalled memory is only ever what you chose to put there.
What it costs
Connecting costs nothing, the free tier is enough to run the first session above, and checking a receipt is always free. For $0.99, a Day Pass covers 24 hours of full use, framed for this page as one full day of connected work across your AIs where the jobs leave records you can show. Ongoing use is the Zambo Pass at $29 a month or $290 a year, and $19 a month for $ZAMBO holders; see pricing for the tiers. Humans pay by PayPal or card through PayPal; agents pay USDC on Base through x402. One account covers every client you connect, so the cost question is asked once, not once per app.
Honest limits
Three edges, stated plainly. First, tools extend what an AI can do, not how well it judges: a connected model can still pick the wrong tool, misread a result, or produce weak work, and your review stays part of the loop. Second, the record covers work run through connected tools: conversation, drafts, and local actions outside tool calls leave no receipt, and honest delivery says so rather than implying full coverage. Third, client support for MCP is still moving: menu names, config formats, and connector flows change between releases, which is why each per client guide on this hub is written against that client current flow and dated, and why the verification loop on this page matters more than any screenshot.
Frequently asked questions
What does it mean to give my AI tools?
It means connecting the AI to an MCP server, a hosted service at one URL that offers named tools the model can call during a conversation or agent session. With Zambo connected at https://zambo.dev/mcp, your AI gains 153 tools in total with 35 on the default surface, plus explicit memory, continuity between AIs, and a verifiable receipt for each saved tool result, instead of being limited to talking about work it cannot run.
What is an MCP server in plain words?
An MCP server is a service that lists tools in a standard format so any MCP aware AI client can use them. MCP, the Model Context Protocol, is the shared wiring: the client, ChatGPT, Claude, Cursor, VS Code, Windsurf, Codex, and others, connects once to the server address, reads its tool list, and can then call those tools in session. One server therefore works across many AIs without being rebuilt for each.
What are the ways to connect tools to my AI?
There are three. Chat based AIs like ChatGPT, Claude, and Grok connect through an in-app connector where you paste the server URL. Editors like Cursor, VS Code, Windsurf, Cline, and Zed connect through a small config file with one entry for the server. Terminal tools like Codex connect with one command, such as codex mcp add, which writes the config for you. All three ways land on the same endpoint and the same account.
Is it safe to connect an MCP server to my AI?
It is as safe as the servers you choose, so choose deliberately. Read any config before pasting it, prefer hosted servers reached by one URL over local code you have not reviewed, and be wary of setups demanding keys a plain connection does not need: Zambo connects with no API key. Keep the list of connected servers short, remove ones you stop using, and save memory deliberately, since memory only holds facts you explicitly tell it to save.
How much does it cost to give my AI tools through Zambo?
Connecting is free and the free tier covers a first working session, and checking a receipt is always free at https://zambo.dev/verify/. For $0.99, a Day Pass covers 24 hours of full use across every AI connected to your account. The Zambo Pass is $29 a month or $290 a year, and $19 a month for $ZAMBO holders. Humans pay by PayPal or card through PayPal; agents pay USDC on Base via x402.