Give your agents a real backend.

Claude, Cursor, Lovable — any MCP client gets filesystem, Git, databases, web search, shell and 50+ more tools through one config entry. Managed projects, live previews, zero infrastructure.

Start building View docs

1Create your workspace
2Generate an API key
3Paste the config — done

mcp-config.json

{
  "mcpServers": {
    "zeeb": {
      "url": "https://backend.zeeb.cloud/mcp/mcp",
      "headers": {
        "Authorization": "Bearer zb_••••••••••••"
      }
    }
  }
}

Paste this into your client. That’s the whole setup.

Works with

ClaudeClaude CodeCursorWindsurfLovableAny MCP client

Use cases

What agents do with a real backend

Stop wiring up local tool servers. Point your agent at Zeeb and let it ship.

Let Claude build entire apps

Claude writes code straight into a managed project — a real Git repo, committed and pushed on every change. You own the history.

zeeb-cloud-projects/9f3a…

Give Cursor a real backend

Query databases, call HTTP APIs, run shell commands and search the web — from inside your editor, with nothing installed locally.

$ run · query · fetch

See it live, instantly

Every project ships with an always-on preview environment that redeploys itself on every commit. Share the URL, get feedback.

https://9f3a….preview.zeeb.cloud

Tool catalog

50+ tools. One mcpServers entry.

Everything below is live the moment your key is created — no plugins, no servers to run.

Filesystem

read_filewrite_filelist_dirsearch_files

Git

commitpushbranchdifflog

Web

web_searchfetch_urlextract

Databases

queryschemamigrate

Shell

run_commandprocess_status

HTTP

requestwebhook

and 30+ more in the docs

write_filewrite_file("src/app.py", code)

Written into the project repo, committed automatically.

web_searchweb_search("fastapi auth best practices")

Fresh results, returned straight to your agent.

run_commandrun_command("pytest -q")

Executed in an isolated worker, output streamed back.

Under the hood

What happens on every tool call

Every call runs the same five-step pipeline — auditable, quota-aware, isolated.

01
Gateway
Resolves your tenant, validates the key or OAuth token.
02
Quota
Billing checks your plan’s limits before anything runs.
03
Queue
An idempotent job is created on the execution plane.
04
Execute
An isolated worker runs the tool and reports back.
05
Settle
The call is settled with billing. One call, one unit.

You are billed exactly here — per tool call. Never per token.

Digital sovereignty

Your data stays in Germany.

No trip across the Atlantic, no reach from foreign authorities. Zeeb runs in German data centres — built and operated by a team in Germany.

Hosted in Germany

Your projects and data live in German data centres at IONOS — under German and European jurisdiction, not someone else’s.

🇩🇪 IONOS · Germany

Built in Germany

Engineered and operated by a team in Germany. German support, German contracts, people you can actually reach.

Made in Germany

GDPR from the ground up

Data protection is the foundation, not an afterthought. No CLOUD Act exposure, no transfer to third countries.

GDPR-native

Your data belongs to you

Every project is a standard Git repo you can clone and take with you at any time. No lock-in, ever.

No lock-in

More than an endpoint

Projects, previews, teams and analytics — the parts you’d otherwise build around your agent.

Git-backed managed projects

Every project is a real repository you own. Clone it, move it, leave anytime — the remote stays authoritative, so there is no lock-in.

Always-on previews

Each project gets a stable preview URL that self-heals and refreshes on every commit.

9f3a….preview.zeeb.cloud

Teams & roles

Isolated workspaces, member roles, and scoped API keys that revoke instantly.

Usage analytics

Live per-tool and per-project usage on your dashboard — you always know where you stand against quota.

Pricing

Priced per tool call. Never per token.

Start free, upgrade when your agents get busy. No hidden fees.

No credit card required on the Free plan. Compare all plans

FAQ

Frequently asked questions

What exactly is Zeeb?

A hosted MCP backend. We run the Model Context Protocol server and 50+ tools — file system, Git, databases, web search and more. You point your AI client at one URL with an API key and your agent can use them immediately.

Which AI clients are supported?

Anything that speaks MCP over HTTP: Claude (Desktop and Claude Code), Cursor, Windsurf, Lovable, or your own agent. If it can load an mcpServers config, it works with Zeeb.

How does pricing work?

Plans are priced on tool calls per month — one call to one tool counts once. We never meter tokens, seats are governed by your plan’s member limit, and you can see live usage on your dashboard at any time.

Is there a free plan?

Yes. Sign up, create a workspace, and start making tool calls — no credit card required. Upgrade only when you need higher quotas, more projects, or more members.

Can my team share a workspace?

Yes. Each workspace (tenant) is fully isolated; invite members, assign roles, and manage scoped API keys per project from the portal.

How is my data secured?

Services talk to each other over mTLS, sessions use short-lived JWT-scoped tokens, API keys are scoped and instantly revocable, and tenant data is isolated at the row level.

Ready to give your agents a backend?

Create your workspace and connect your first MCP client in under five minutes.

Start building View docs