Everything your AI should know, behind one secure link.
One MCP server for your personal or team use — your documents, apps and memory behind a single link, so you don't wire your assistants to hundreds of servers. You approve what AI can see and do, and every call is logged.
Free tier — 1 server, 1,000 requests a month. No credit card. Export or delete everything at any time.
Works with every MCP client
Cursor
Windsurf
ZedReady-made for the tools you run
HubSpot
Slack
Gmail
Calendar
NotionYou've taught Claude everything. ChatGPT starts from zero.
Your business, your clients, your terms — explained again in every new chat, to every new model.
Everything you taught Claude stays with Claude. Move to ChatGPT and you're a stranger again.
An answer without its source is a guess you have to verify yourself.
Even when it's right, you still copy it into email, calendar and CRM by hand.
Your AI's brain shouldn't belong to one vendor. Build one shared brain, point every assistant at it, and keep the keys.
— From the Protobox team
One shared brain.
Every AI plugged into it.
Actions from your apps, knowledge from your docs, prompts from your playbook, memory that follows you.
Learn moreYou decide exactly what AI can see and do — per-action toolsets, an audit log of every call, one-click revoke.
Learn moreOne secure link that works in Claude, ChatGPT, Cursor — and anything else that speaks MCP.
Learn moreEverything your AI needs
Tools, knowledge, prompts, memoryMost platforms stop at tools. MCP defines three primitives — tools, resources, prompts — and Protobox hosts all three from one endpoint, with credentials in an encrypted vault, never in a prompt.
You decide exactly what AI can see and do
Toolsets, audit log, one-click revokeBuild toolsets — scoped bundles of just the actions a job needs. Read-only for the research agent. Create-but-never-delete for support. Every call is logged, and revoking takes one click. Coming next: approval gates for sensitive actions, workspace roles, and read/write/destructive tiers.
One URL. Every AI your team uses.
Set it up once — not per app, per person, per assistantYour workspace gets a single secure endpoint. Paste it into Claude, ChatGPT, Cursor — anything that speaks MCP — and the whole team's AI is connected. Sessions are scoped to a toolset, and you can revoke any of them in one click.
Model Context Protocol, in plain English
The standard plug between LLMs and your tools. Open spec, every client, no glue code.
JSON-RPC. Open. Transport-agnostic. Anthropic shipped it; current revision 2025-11-25.
Claude, ChatGPT, Cursor, Windsurf, Zed — plus every serious agent framework.
Hosting, OAuth, rate limits, audit logs. Stable URL. Every MCP client connects.
Add your documents, apps and tools, paste one link, ask from anywhere.
Setup is three moves, not a migration. Protobox sits behind the assistants you already open every day.
Add what your AI should know
Drop PDFs, paste a URL, connect apps from the catalog. Sign in, approve, done — synced sources stay where they live.
Paste one link into every assistant
Claude, ChatGPT, Cursor — the same address works in all of them. Add it once; there is nothing else to configure.
Ask anywhere, get the same answer
Every assistant answers from the same shared brain and shows its source. When you're ready, let it act — you choose which actions it can take.
Everything around the work, handled
The plumbing an AI-equipped company needs — already built, already tested.
Custom tools
TypeScript, Python, or an OpenAPI spec — your own actions next to the catalog.
Scoped sessions
Hand out session links that expose a single toolset. Revoke any of them, any time.
Audit log
Every tool call recorded — actor, payload, latency, result.
Secrets vault
OAuth tokens and API keys encrypted at rest, never logged, never in a prompt.
Version history
Integrations are versioned. See exactly which actions changed in every release, down to the tool.
Rate limits & quotas
Request throttling plus monthly quotas, so a runaway agent loop can't burn your stack down.
Workspaces
Separate departments or clients — each with its own endpoint, members, and billing.
Playground
Try any action with real inputs yourself, before an AI ever calls it.
From the blog
Building, shipping, and operating MCP servers.
Dev NotesHow to manage a hosted MCP server from the terminal
Every dashboard click is an ops task you cannot script. Run your hosted MCP server from the terminal instead: tools, toolsets, keys, knowledge, and CI.
Security & Compliance71% of organizations aren't prepared to secure their AI agents' tools
MCP gives AI agents autonomous access to real systems — and introduces attack vectors that traditional security can't see. A technical breakdown of tool poisoning, rug pulls, cross-server shadowing, and the defense framework production teams need now.
Dev NotesHow to register MCP tools without running a server
Hand-rolling an MCP server means owning transport, auth, hosting, and schema drift. Registering the same tools against a hosted MCP endpoint skips all four.
MCP changelog and dev notes
Short, occasional updates on the MCP spec, new Protobox features, and patterns we've seen in production. No marketing fluff.
Questions teams ask before they roll this out
Make the link. Paste it once. Ask from anywhere.
Free to start — one server, a thousand requests a month, no credit card.