Pay Acme Design invoice #2041 for $1,200?
Over your $5 auto-approve limit and money can't be pulled back once sent.
- Agent
- ops-bot
- Amount
- $1,200.00
- Reversible
- No
Cloud for Botsinfrastructure for your bots
Your agent asks you
Needs you
Your agent asks you
only when it matters.
Set rules once. Small, reversible actions under your limits go through. Everything else becomes one plain-English card: Approve or Reject.
autoRenew domain example.com · $3.99under $5 · reversible
autoRetry failed webhook · $0reversible
autoTop up API credits · $4.50under $5
Sample: Watch-only: your rules WOULD auto-approve this. Nothing is approved in free mode; ask your user as usual. Logged for the weekly report.
Works with the agents you already run
- AClaude
- {}Codex
- CChatGPT
- GGrok Bot
How it works
Rules once. Cards only when needed.
set_rulesSay what is fine without asking: a dollar limit, reversible only, allowed kinds of action.
check_actionBefore anything consequential, your agent asks. Answer: auto_approve, needs_approval or deny.
get_decisionFor needs_approval you get a hosted card. Tap Approve or Reject; your agent picks it up.
Pricing
Watch free. Pay when it acts.
Free resets daily and is not a trial. Pro is $9 a month.
Free · watch-only
$0
Watch-only mode. 200 checks a day: logs what your rules would auto-approve, ask about or block, and drafts a weekly report. Never approves anything.
- Logs what your rules would decide
- Approve / ask / block, recorded
- Weekly report of asks you could skip
- Never approves anything
Pro
$9/month
Acting: auto-approvals are binding, and everything else gets a hosted approval card your agent waits on.
- Auto-approvals are binding
- Hosted Approve / Reject card your agent waits on
- 20,000 checks a day
- Key instantly after checkout, cancel anytime
Worked example
Your agent renews 30 small subscriptions and domains a month under $5 each. In free mode the weekly report shows 30 asks your rules would have cleared. With Pro at $9/month, those 30 go through on their own and you only see the 4 that need you.
For agents
Tools your agent gets
Each result includes understood, job and a one-line outcome your agent can report back in plain words.
check_actionBefore a consequential action, ask whether the user's rules auto-approve it. Returns auto_approve, needs_approval (with a plain-English approval card and a link the user can tap) or deny.
get_decisionCheck whether the user approved or rejected an approval card (by card_id and card_token). Only act when status is approved.
set_rulesSave the user's rules (needs an API key so they stay yours). Each rule: {name, when:{type:[...], max_usd, min_usd, reversible, target_ends_with:[...]}, verdict:"auto"|"ask"|"deny"}.
get_rulesShow the rules in effect for your key (or the default starter rules).
weekly_reportDraft a plain-English report of the last 7 days: actions checked, how many your rules would auto-approve, how many needed you, top kinds of asks, and rule suggestions. Read-only; sends nothing.
usage_and_upgradeUsage today + upgrade link (not metered).
Connect
One line to install
claude mcp add --transport http approval-rules https://approve.cloudforbots.com/mcp
MCP URL https://approve.cloudforbots.com/mcp
REST POST https://approve.cloudforbots.com/api/run/<tool>
Auth optional Authorization: Bearer <key>
Codex, ChatGPT, Grok Bot: add a remote MCP server with the URL above. Skill file: SKILL.md · catalog entry: tool.json
FAQ
Questions
What problem does this solve?
Agents either ask about everything, so people rubber-stamp, or ask about nothing, so mistakes slip through. Approval Rules lets your user say once what is fine (reads, drafts, small reversible spend) and what always needs them (deletes, messages to new people, big payments).
How does my agent use it?
Before a consequential action, call check_action with the type, a one-line summary, the amount and target. You get auto_approve, deny, or needs_approval with a plain-English card and a link your user taps Approve or Reject on. Call get_decision before acting.
What is watch-only mode?
The free plan never approves anything. It tells your agent what the rules would decide, logs it, and weekly_report drafts a summary: how many asks the rules would have cleared, how many still need you, and which rules to add. Your agent keeps asking you as usual.
What are the default rules?
Auto-approve reads, searches and drafts, and reversible spend up to $5. Ask about deletes and messages to people. Block anything $500 or more. With an API key, set_rules replaces them with your own, after a preview your user confirms.
Is there a model inside?
No. Bring your own model. Approval Rules only matches the action your agent describes against the rules; it never performs the action, sends messages or moves money.
What does it cost?
Free: watch-only, 200 checks a day, weekly report. Pro is $9 a month: binding auto-approvals, hosted approval cards, 20,000 checks a day. You get your Pro key on screen right after paying.