NEWStripe Connect for paid MCP servers just shipped

The hosted runtime
for MCP servers.

Deploy, authenticate, observe, and monetize an MCP server in one command. We run the edge, handle OAuth for every client, and wire Stripe Connect for you.

Deploy your first server Read the docs

free forever tier · no card required · deploy in 30s

~ / notion-mcp

➜~/notion-mcp

edge · iad1

Live · right now

updating every second

Tool calls · today

1,842,930

▲ +18.2% vs yesterday

Servers

12,847

Edge p95

78ms

Global edge

22 regions · every request < 100ms

anycast

iad1sfo1lhr1cdg1fra1bom1sin1syd1+14 more

30-second deploys

Global anycast edge in 22 regions. Zero cold starts.

OAuth, everywhere

Claude, Cursor, Windsurf, Cline, Continue, Zed.

Live observability

p95 latency, error rate, per-tool audit — free.

Developer experience

Write your server.
We ship it.

No Dockerfiles. No Terraform. No YAML. Write a TypeScript or Python server that looks like ten normal functions — we handle the rest.

notion-mcp.ts

import { createMcpServer } from "@cognoverge/mcp";

export default createMcpServer({
  name: "notion-mcp",
  tools: {
    search: {
      description: "Search a Notion workspace",
      input: { query: "string" },
      handler: async ({ query }, { user }) => {
        const results = await notion.search(query, user.token);
        return results.map(page => ({
          title: page.title,
          url: page.url,
          preview: page.preview,
        }));
      },
    },
    create: {
      description: "Create a new Notion page",
      input: { parent: "string", title: "string" },
      handler: async ({ parent, title }, { user }) => {
        return notion.pages.create({ parent, title }, user.token);
      },
    },
  },
});

39 lines · ready to deploySee more templates

cognoverge · deploy

➜cognoverge deploy

› Reading ./notion-mcp.ts

✓Bundled 5.2 kB · 380ms

✓Provisioning edge deployment...

✓Deployed to https://notion.devesh.cognoverge.dev

✓OAuth configured · Google GitHub magic-link

✓Rate limits, quotas, audit log · online

✓Edge regions iad1 sfo1 lhr1 +19

✓ Live in 23.4 seconds

Next: add to Claude Desktop

➜npx cognoverge add notion --client=claude

● Live · streaming logsedge · iad1

The product

Everything an MCP server needs.
Zero rebuilds.

Four surfaces, one platform. Use any one alone — they compose without configuration.

Gateway

Global edge in 22 regions.

Push your server, get an HTTPS endpoint on a global anycast edge. Custom domains, branch previews, versioned rollback. Zero cold starts.

iad1sfo1lhr1cdg1fra1bom1sin1syd1hnd1gru1icn1dxb1+ 10 more

Auth

OAuth for every MCP client.

Google, GitHub, magic-link out of the box. BYO OIDC on Team.

Claude Desktop

Cursor

Windsurf

Cline

ChatGPT Desktop

Observability

Calls, latency, errors — free.

p50/p95/p99 per server, per tool, per client — with full request audit trail.

p50

42ms

p95

78ms

p99

156ms

Monetize · 5% take rate

Charge per call. We handle Stripe.

Stripe Connect onboarding, KYC, tax, and payouts wired up. Weekly deposits to your bank — net of 5%.

Revenue · 30d

$4,218

▲ +22% vs last month

Next payout

$1,596

Mon · ICICI ••5482

Read the full docs

The dashboard

One pane of glass.
Every server, every client.

Every metric that matters, in the same tab. No Datadog invoice, no Grafana bolted on the side.

app.cognoverge.com/workspace/devesh/overview

⌘K

Overview · last 30 days

All servers · all clients · all regions

healthy

Calls · 24h

817,924

▲ +18.2%

Edge p95

78ms

▲ −4ms

Error rate

0.23%

▲ −0.04%

Revenue · MTD

$4,218

▲ +22.4%

Calls per hour

30 hours · UTC

tools/calltools/list

LIVE CALLS

allerrorsslow

21:04:12tools/callnotion-mcp · searchClaude · Sarah K.142ms200

21:04:08tools/callgithub-mcp · pr.listCursor · Devesh R.218ms200

21:04:03tools/listnotion-mcpClaude · Sarah K.38ms200

21:03:58tools/callnotion-mcp · createWindsurf · alice@312ms200

21:03:51tools/calllinear-mcp · ticket.createCursor · Devesh R.198ms429

21:03:42tools/callgithub-mcp · file.readCline · jose@88ms200

Observability, without Datadog

Every call. Every span.
Full-fidelity by default.

Every MCP call captured with the full request body, tool arguments, downstream span tree, and OAuth identity — retained for 30 days on Pro, 90 days on Team. No sampling. No agent to install.

TRACE · trc_a4f2ee12 312ms · 9 spans · 21:04:12 UTC

Servernotion-mcp

Toolsearch

Caller[email protected]

ClientClaude Desktop

Regioniad1

Status200 OK

0ms80ms160ms240ms312ms

notion.search+28ms · 218ms

api.notion.com · POST /v1/search

{
  "method": "POST",
  "url": "https://api.notion.com/v1/search",
  "body": { "query": "Q3 roadmap", "page_size": 20 },
  "response": { "status": 200, "results": 12, "duration_ms": 218 }
}

Alerts

Alert on anything.

Rule · 07active

When p95 latency > 200ms for 2 minutes on notion-mcp

notify #ops-mcp · Slack

Rule · 04active

When error rate > 1% across any server

page oncall · PagerDuty

Anomaly detected

error rate 0.2% → 1.4% on stripe-mcp

detected · 21:01 UTC · auto

LIVE TAIL

server:notion-mcp code:!=200

allerrorsslow > 300msadd filter

TimeServer · toolClient · identity mscode

Request429 Too Many Requests

linear-mcp · ticket.create

trc_a4f2ee12 · 21:04:10.732 · 198ms

{
  "arguments": {
    "title": "Q3 planning",
    "assignee": "[email protected]",
    "priority": "high"
  },
  "identity": "[email protected]",
  "oauth": "github",
  "region": "iad1",
  "quota": "78 / 100 req·min"
}

Rate limit hit. Caller exceeded 100 req·min. Cognoverge returned 429 with a Retry-After header. Bump the quota in Settings → Quotas.

Streaming · 24 events in the last 60s30-day retention·export → S3 / BigQuery

99.99%

Uptime SLO

24 months rolling

78ms

Global p95 latency

22 edge regions

12k+

MCP servers running

in Cognoverge today

$0.001

Cost per tool call

at scale · Pro tier

Devs shipping on Cognoverge

The infra layer for the next wave of AI tools.

“We had four devs writing Docker files and OAuth wiring for our internal MCP server. Cognoverge replaced two weeks of glue with a single deploy command.”

Priya Anand

Staff Engineer · Ranger AI

“Stripe Connect out of the box was the killer feature. We turned a weekend project into a $3K/mo side business without touching billing code.”

Marcus Chen

Indie dev · SiteGPT · @marcuschen

“The OAuth-per-MCP-client conformance is the part nobody talks about but everyone needs. Cognoverge handles it and it just works in Claude, Cursor, and Windsurf.”

Elena Vasquez

Head of Platform · Loopstack

Teams building on Cognoverge

Ranger AILoopstackKettlecorn LabsFieldenNorthpathStudiokitHalcyonDuofold

Pricing

Free to start.
Cheap to grow.

No hidden compute charges. Stripe Connect is opt-in. Cancel anytime.

Free

forever · no card

Ship your first MCP server people can actually use.

1 MCP server
10,000 calls / month
Community OAuth (Google + GitHub)
Real-time logs · 7-day retention
Community Discord

Deploy free

Quick answers.

Model Context Protocol (MCP) is Anthropic's open standard for giving AI clients — Claude Desktop, Cursor, Windsurf, and others — access to your tools and data. An MCP server exposes tools (like search Notion, create a Linear ticket) that any MCP-capable client can call. Cognoverge hosts these servers in production.

Stop yak-shaving.
Start shipping.

One command. Production HTTPS, OAuth, observability, and optional billing. Free tier covers your first 10K calls.

Open my workspace Read the docs

30-second deploys · no card · cancel anytime

loading workspace…

import { createMcpServer } from "@cognoverge/mcp"; export default createMcpServer({ name: "notion-mcp", tools: { search: { description: "Search a Notion workspace", input: { query: "string" }, handler: async ({ query }, { user }) => { const results = await notion.search(query, user.token); return results.map(page => ({ title: page.title, url: page.url, preview: page.preview, })); }, }, create: { description: "Create a new Notion page", input: { parent: "string", title: "string" }, handler: async ({ parent, title }, { user }) => { return notion.pages.create({ parent, title }, user.token); }, }, }, });

{ "arguments": { "title": "Q3 planning", "assignee": "[email protected]", "priority": "high" }, "identity": "[email protected]", "oauth": "github", "region": "iad1", "quota": "78 / 100 req·min" }

The hosted runtimefor MCP servers.

Write your server.We ship it.

Everything an MCP server needs.Zero rebuilds.

Global edge in 22 regions.

OAuth for every MCP client.

Calls, latency, errors — free.

Charge per call. We handle Stripe.

One pane of glass.Every server, every client.

Every call. Every span.Full-fidelity by default.

Alert on anything.

The infra layer for the next wave of AI tools.

Free to start.Cheap to grow.

Quick answers.

Stop yak-shaving.Start shipping.

The hosted runtimefor MCP servers.

Write your server.We ship it.

Everything an MCP server needs.Zero rebuilds.

Global edge in 22 regions.

OAuth for every MCP client.

Calls, latency, errors — free.

Charge per call. We handle Stripe.

One pane of glass.Every server, every client.

Every call. Every span.Full-fidelity by default.

Alert on anything.

The infra layer for the next wave of AI tools.

Free to start.Cheap to grow.

Quick answers.

Stop yak-shaving.Start shipping.

The hosted runtime
for MCP servers.

Write your server.
We ship it.

Everything an MCP server needs.
Zero rebuilds.

One pane of glass.
Every server, every client.

Every call. Every span.
Full-fidelity by default.

Free to start.
Cheap to grow.

Stop yak-shaving.
Start shipping.

The hosted runtime
for MCP servers.

Write your server.
We ship it.

Everything an MCP server needs.
Zero rebuilds.

One pane of glass.
Every server, every client.

Every call. Every span.
Full-fidelity by default.

Free to start.
Cheap to grow.

Stop yak-shaving.
Start shipping.