TaskLite × Gemini

A real backend and admin,
from a Gemini conversation.

TaskLite is an MCP server Gemini CLI can install as an extension in one command. Describe the data and Gemini creates projects, boards with typed columns, validation rules, automations and REST endpoints, and the business gets an admin it can operate. The Gemini API gets the same tools through the google-genai SDK.

Install the TaskLite extension in Gemini CLIOAuth sign-in, nothing to configure
  1. 1Run the install command below. The extension registers the hosted server with OAuth discovery.
  2. 2Start Gemini CLI, type /mcp: the first call opens your browser to sign in to TaskLite (or sign up on the spot).
  3. 3Say what you need: "create a project for my repair shop with customers, jobs and parts, then expose jobs as an endpoint".
Or add the server by URL:
https://mcp.tasklite.net/mcp

Without the extension, put the server in ~/.gemini/settings.json with httpUrl and authProviderType "dynamic_discovery".

Commands and SDK
Gemini CLI extension
gemini extensions install https://github.com/shimon-ks/tasklite-mcp
settings.json, with an API key instead of OAuth (scripts, shared machines)
{
  "mcpServers": {
    "tasklite": {
      "httpUrl": "https://mcp.tasklite.net/mcp",
      "headers": { "Authorization": "Bearer tl_your_key" }
    }
  }
}
Gemini API: pass the MCP session as a tool
from google import genai
from mcp import ClientSession
from mcp.client.streamable_http import streamablehttp_client

client = genai.Client()
async with streamablehttp_client(
    "https://mcp.tasklite.net/mcp",
    headers={"Authorization": "Bearer tl_your_key"},
) as (r, w, _):
    async with ClientSession(r, w) as session:
        await session.initialize()
        resp = await client.aio.models.generate_content(
            model="gemini-2.5-pro",
            contents="Which orders are overdue, and who owns them?",
            config={"tools": [session]},
        )
        print(resp.text)
Start free

How it works

1. Describe the business

Gemini picks typed columns by meaning: dates, phones, currency, dropdowns, relations. The server refuses an obvious mismatch and suggests the right type, so the model self-corrects.

2. Gemini builds and wires it

One build_backend call: project, boards, validation rules, relations, sample rows, REST endpoints and an API key. Then a public form and automations with delays and retries. get_frontend_prompt hands the next agent everything it needs.

3. The business runs it

Seven views, forms, portals, approvals and an audit log in app.tasklite.net. The client changes a field without calling you; your endpoints keep working through aliases.

Built for both of you

Developers on Gemini CLI

The extension carries a GEMINI.md context file, so the model already knows the build order, the column types and how to read a refusal. Your first backend takes about a minute.

Teams building on the Gemini API

One API key, the session passed as a tool, and Gemini answers from and writes to your live boards. Row-level security through X-App-User when your app has its own users.

Scoped and revocable

Gemini acts with your permissions through short-lived OAuth tokens, or through a personal API key you can revoke from Integrations. Every write is in the audit log. Data is hosted in the EU.

Does the extension work on Windows?
Yes. Gemini CLI installs extensions from a Git URL on every platform; the server is hosted, so nothing runs locally except the CLI.
Can I use the local package instead of the hosted server?
Yes: npm install -g @tasklite/mcp, then a stdio entry in settings.json with command "tasklite-mcp" and TASKLITE_API_KEY in env.
Which Gemini models can call tools?
Any model that supports function calling in the API, including Gemini 2.5 Pro and Flash. In Gemini CLI, tools are available with the default model.

Install TaskLite in Gemini CLI

Free forever for 2 operators, no card. Team $5 per operator per month when the team joins.

Create your TaskLite account