MODEL CONTEXT PROTOCOL

Connect Claude. Ship from chat.

QuikRun's MCP server gives any MCP client a set of tools to create, run, and deploy snippets, then hand back a live URL without leaving your editor.

Start free

MCP is on the free tier. Mint a token in the dashboard.

What you get

MCP (Model Context Protocol) is an open standard that lets AI clients call external tools. QuikRun implements a remote MCP server so your agent can build and deploy real endpoints for you.

You describe the outcome in chat; the agent calls QuikRun tools to write the code, run it, deploy it to the edge, and return the URL, all reviewable before it ships.

Add it to your client

One HTTP endpoint, every MCP client. Add your token as a bearer header, then pick yours:

Claude
claude mcp add --transport http quikrun https://mcp.quik.run --header "Authorization: Bearer quik_..."
Cursor
// .cursor/mcp.json
{
  "mcpServers": {
    "quikrun": {
      "url": "https://mcp.quik.run",
      "headers": { "Authorization": "Bearer quik_..." }
    }
  }
}
VS Code
code --add-mcp '{"name":"quikrun","url":"https://mcp.quik.run","headers":{"Authorization":"Bearer quik_..."}}'
Windsurf
// ~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "quikrun": {
      "serverUrl": "https://mcp.quik.run",
      "headers": { "Authorization": "Bearer quik_..." }
    }
  }
}
Zed
// settings.json → context_servers
"quikrun": {
  "source": "custom",
  "url": "https://mcp.quik.run",
  "headers": { "Authorization": "Bearer quik_..." }
}

Ask your agent

Once QuikRun is connected, say things like:

›Create a snippet that returns today's F1 standings as JSON, then deploy it.
›Build a webhook that posts failed Stripe charges to #payments in Slack.
›Show me my snippets and open the one called refund-alerts.
›Run refund-alerts with a sample Stripe payload and show me the output.
›Make refund-alerts public and give it a friendlier name.
›Duplicate my og-image snippet so I can tweak a copy.

Tools your agent gets

list_snippets
() → { snippets }

Lists your snippets with slug, language, visibility, endpoint, and run count.

get_snippet
(slug) → { code, meta }

Fetches a snippet's metadata and full source so the agent can read or edit it.

create_snippet
(name?, language?, prompt?) → { slug }

Creates a snippet, optionally from a name, language, and description.

update_snippet_code
(slug, code) → { diff }

Saves source code to the snippet's draft and returns the diff.

run_snippet
(slug, method?, body?, …) → { output, logs }

Runs a snippet and returns its output, HTTP response summary, and logs.

deploy_snippet
(slug) → { url }

Publishes the current draft as the live version at its URL.

update_snippet
(slug, name?, visibility?, …) → { ok }

Updates settings: name, visibility, language, or runtime.

duplicate_snippet
(slug) → { slug }

Duplicates a snippet so you can fork and tweak a copy.

delete_snippet
(slug) → { ok }

Permanently deletes a snippet. Destructive and cannot be undone.

In practice

YOU

Create a snippet that returns today's F1 standings as JSON, then deploy it.

CLAUDE
▸ quikrun.create_snippet · f1-standings: ok · 0.9 s
▸ quikrun.run_snippet · 200 OK · 48ms
▸ quikrun.deploy_snippet · live

Done. f1-standings fetches the current standings and returns them as JSON. I ran it once to check the output, then deployed. It's live now:

quik.run/r/f1-standings

Give your agent a runtime.

Connect QuikRun's MCP server and ship from any chat. The first prompt is free.

Start freeRead the MCP docs