Start free
model context protocol

The TinyCommand
MCP server

coming soon

Describe a goal in your AI chat and TinyCommand builds the system and runs it, without leaving Claude, Cursor, or ChatGPT. Your forms, tables, workflows, agents and email, reachable over one open protocol.

get notified what is MCP?
mcp.tinycommand.com/v114 toolsstreamable http
claude ▸ tinycommand
you▸ capture leads from a form, score them, email the hot ones
tc▸ build_system
planning form → workflow → agent → table → email
✓ 5 assets ready in your workspace
you▸ how many scored above 80 this week?
tc▸ query_records
› 42 rows · score ≥ 80

Most AI chats can talk about your work but cannot touch it. The Model Context Protocol is the open standard that lets an assistant reach real tools and data. TinyCommand is putting its whole platform behind it, so the AI you already use can build and run your back office for you.

01 / capabilities

14 tools, shaped around real jobs

Not raw API endpoints. Journey-shaped tools that build new systems, operate your live workspace, and discover what is already there.

tinycommand-mcp ▸ tools/list14 tools
build4
build_systemturn a described goal into a working, multi-asset system
update_assetchange an existing form, table, workflow or agent
confirm_savesave the built system into your workspace
abort_buildstop a build that is still in flight
operate8
query_recordsread rows from one of your TinyTables
add_recordadd a row
update_recordupdate a row
trigger_workflowkick off a workflow
send_via_integrationsend through your connected Gmail, Slack and more
run_agentrun one of your agents
run_enrichmentenrich a company or a person
check_enrichmentcheck on an enrichment job
discover2
list_my_assetslist your forms, tables, workflows and agents
describe_my_workspacedescribe the workspace and what it can do
02 / connect

Works in the AI chat you already use

Standard MCP over an HTTP endpoint. Any compliant client connects with one config block. Add it, and the tools show up.

Claude DesktopCursorcustom GPTsany MCP client
~/.config/mcp.json
{
  "mcpServers": {
    "tinycommand": {
      "url": "https://mcp.tinycommand.com/v1",
      "headers": { "Authorization": "Bearer tc_sk_..." }
    }
  }
}
03 / setup

Three steps, no SDK

  1. 01generate a workspace keyscope it to read, build, write or execute, then connect. revoke any time.
  2. 02paste one config blockdrop the url and key into Claude, Cursor, or a custom GPT. no SDK to install.
  3. 03talk to your workspaceask your AI to build a system or run an action. it happens inside TinyCommand.

Keys are workspace-scoped with per-key permissions, every call is audited, and you can revoke a key at any time. Start read-only, grant more as you trust it.

04 / waitlist

Be first when it goes live

The server and its tools are built. We are finishing authentication and rollout now. Leave your email for one message the day it opens, plus the setup guide.

05 / faq

Questions

Is the TinyCommand MCP server live yet?
Not yet, it is coming soon. The server and its 14 tools are built and running at mcp.tinycommand.com/v1, and we are finishing authentication and rollout before opening it up. Leave your email and we will tell you the day it goes live.
Which AI clients will it work with?
Any MCP client. At launch that means Claude Desktop, Cursor, and custom GPTs, with more clients adopting the protocol every month. You connect with a single configuration block, no SDK to install.
What will I be able to do with it?
Build: describe a system in chat and TinyCommand assembles the forms, tables, workflows, agents and email. Operate: query and update records, trigger workflows, send through your connected integrations, run agents, and run enrichment. 14 tools in total, shaped around real jobs rather than raw API calls.
How is it secured?
Every connection uses a workspace-scoped API key with per-key permission scopes, so a key can be read-only or allowed to build and write. Keys are revocable at any time, every tool call is audited, and requests are rate limited.
Will I need to write code?
No. You generate a key and paste one config block into your AI client. Building agents inside TinyCommand needs no code either. MCP is simply the bridge for reaching that same workspace from outside tools.
How is this different from building an agent inside TinyCommand?
Building inside TinyCommand puts the agent next to your data on one canvas. The MCP server is the reverse direction: it lets an outside agent in Claude, Cursor or ChatGPT reach that same workspace and run the same actions. You can use both.

Do not want to wait for the protocol?

You do not need MCP to put an AI agent to work on your data. Build one inside TinyCommand today, next to your forms, tables and email, and connect it over MCP later.

Start free See TinyAgents