lidfly.ai
Dashboard

AI-first advertising control plane

Manage every ad channel by talking to your AI.

LidFly is being built to connect Google Ads and other advertising platforms to the assistant you already use. Provider integrations are in development, with production access subject to each platform's approval.

  • Google Ads integration in development
  • Every change is approved
  • Complete action log

Planned provider channels share one control plane, with every client account kept inside an explicit workspace scope.

Channels

Four systems, one set of commands.

Each planned channel gets a normalized adapter with the same operations for accounts, statistics, campaigns, and creatives. Your assistant learns one contract instead of four different APIs.

GAintegration in development

Google Ads

Search, Performance Max, and shopping campaigns with account and keyword statistics.

7 operations

MEintegration in development

Meta Ads

Facebook and Instagram in a single adapter: campaigns, creatives, and audiences.

8 operations

TTintegration in development

TikTok Ads

Video campaigns, creatives, and delivery statistics down to the ad group.

7 operations

TGintegration in development

Telegram Ads

Channel placements, budgets, and results inside the Telegram Ads account.

6 operations

Next in lineMicrosoft AdsLinkedIn AdsPinterest AdsEtsy Ads

Two ways to work

Bring your own assistant, or use ours.

You already pay for an AI subscription, so we charge for access and methodology instead of tokens. If you would rather not set anything up, the built-in chat covers the model for you.

MCP connection

Your own assistant

Connect LidFly as an MCP server to Claude, ChatGPT, or Cursor and work where you already think and write.

  • One address and one key for every channel
  • Tools, resources, and methodologies arrive with the connection
  • Your assistant subscription pays for the model
in development

Built-in chat

The dashboard will host the same tools in a chat where the model is on us, so a new account is productive in a minute.

  • Nothing to install or configure
  • The same safety boundaries as MCP
  • Shared history for the whole team

Both paths run through one control plane: the same tools, the same limits, the same log.

How it works

The planned account workflow in four steps.

  1. 01

    Connect an account

    When an integration is available, the user initiates OAuth. Provider credentials stay in encrypted secret storage and never reach the model.

  2. 02

    Take your key

    The dashboard issues one MCP address and one key for your assistant. Revoke it whenever you want.

  3. 03

    Ask your question

    Before any API call the assistant receives the capability map, the account scope, and the platform limits.

  4. 04

    Approve the change

    A write waits for your confirmation, carries an idempotency key, and lands in an append-only log.

Planned workflows

Questions instead of dashboards.

Examples the provider adapters are designed to support after connection. Reads stay within the authorized account scope, and campaign changes stop for approval.

Which campaigns lost ROAS last week, and why?

Compares two periods and names the segments behind the drop.

Show tomorrow's budget plan across all channels.

Collects budgets from every connected account into one table.

Pause everything above the target cost per action.

Prepares the list and waits for approval before a single change.

Which creatives burned out in Meta this month?

Tracks frequency and the decline in click-through rate by creative.

Put together the weekly report for the client.

Gathers results from four systems and explains what changed.

Why did spend jump on Saturday night?

Finds the anomaly by hour and links it to the change that caused it.

Safety

An assistant with brakes.

An AI with access to a live budget needs boundaries, not good intentions. They are built into the architecture, not promised in a policy document.

  • 01

    Reading is free, writing is not

    Read operations run immediately. Budget, campaign, audience, and creative changes require a declared risk level and an approval policy.

  • 02

    A repeat does not double the spend

    Every write carries an idempotency key, so a retry after a network error updates the same operation instead of creating a second campaign.

  • 03

    An audit trail you cannot edit

    Who, what, when, and with which result — every action is appended to a log that the application itself cannot rewrite.

  • 04

    Secrets stay outside the model

    Provider credentials live in encrypted storage. Tokens, authorization headers, and customer lists never enter logs or the assistant context.

Methodologies

We hand the assistant knowledge, not only an API.

The versioned skill catalog travels through the same MCP connection: how each platform behaves, what it forbids, and which mistakes cost money.

advertising-write-safety

Safe campaign changes

The mandatory procedure before touching a budget, a bid, or a status: what to check, what to confirm, what to record.

google-ads-account-discovery

Choosing the right account

How to connect Google Ads, tell a manager account from a client one, and avoid working in the wrong place.

skills/list · sha256 digest

Versions and access

Each methodology has a content digest and a visibility level, and every activation is written to the ledger.

Point your AI at your advertising.

Sign in with an email code, and your workspace is created on the first visit. Connecting an advertising account is the next step.

Open the dashboard

Passwordless sign-in · no card required