For AI power users
Your agent does the work.You approve.
One MCP server for your AI agent: it drafts the change, Taseli checks it, you approve.
- Free
- Nothing changes without your approval

Why did cost per lead go up this week?
Google broadened matching, and 31% of spend moved to generic terms. I prepared 6 negatives.
Read-only is too little.Writing with no control is too much.
“My agent tells me what's wrong, then I copy the fix in by hand.”
“Some connectors change budgets with no caps and no log.”
“I don't want an agent approving its own changes.”
Your agent does the legwork. Taseli keeps it safe.
The agent analyzes and drafts. Every change still passes caps, a dry run and the log.
One MCP server, every client
Claude, ChatGPT, Cursor, Gemini and other MCP clients, with OAuth sign-in.
Proposals, never direct writes
Every write returns an approval link. No agent approves itself.
Caps and a dry run on every change
Caps are enforced on our server, and each change is tested first.
The proposal shows up in chat. You approve in Taseli.
The approval is tied to the exact change. If it changes, you approve again.
Lower the daily budget and pause ads overnight
| - | Current | Proposed | |
|---|---|---|---|
| Daily budget | $140 | $105 | |
| Ad schedule | All day | 6 am – 11 pm |
$730saved a month (est.)
Frequently asked questions
Doesn't an approval step slow my agent down?
Which MCP clients work?
Can my agent approve its own proposals?
What happens if my agent is fed a malicious instruction?
The plan that fits
Pro
Pro includes MCP reads for three accounts; Power/MCP ($29 a month, billed annually) adds agent proposals and GAQL.
- Everything in Starter
- Runs on its own within limits you set
- Checks every hour
$99/month
Billed annually · 30-day refund
Check my account for free- Free
- Nothing changes without your approval