Run Budget Guard
by botrecipe.com · remote · no sign-in to list tools · v1.0.1 · published by com.botrecipe
Hard dollar, token, call and time caps shared by every sub-agent in one agent run.
6 tools: 4 change things and 2 only read.
Scanned Oct 21, 2026
- Deletes or cancels0Claude Code: deny
- Can't be undone0Claude Code: ask
- Changes things4Claude Code: ask: see these tools
- Only reads2Claude Code: allow: see these tools
3 checks matched something in the tool definitions. No finding is not a safety guarantee. How the checks work.
claude mcp add --transport http budget https://budget.cloudforbots.com/mcpFindings
Matches of fixed checks, with the text they matched.
- Highreserve
The input schema at properties.handle.description contains "<secret>", a tag of the kind used to mark instructions for the model.
What this check looks forMatched text
inputSchema.properties.handle.description: "...le from open_budget (looks like bg_<id>.<secret>). Pass it to every sub-agent so they al..."
- Highrecord
The input schema at properties.handle.description contains "<secret>", a tag of the kind used to mark instructions for the model.
What this check looks forMatched text
inputSchema.properties.handle.description: "...le from open_budget (looks like bg_<id>.<secret>). Pass it to every sub-agent so they al..."
- Highbudget_status
The input schema at properties.handle.description contains "<secret>", a tag of the kind used to mark instructions for the model.
What this check looks forMatched text
inputSchema.properties.handle.description: "...le from open_budget (looks like bg_<id>.<secret>). Pass it to every sub-agent so they al..."
Tools
Open a tool for its reasons and full description. Descriptions are the publisher's own words, shown as they are.
6 tools
Changes things 4Claude Code: ask
open_budgetOpen a hard budget for one agent run before it starts: max dollars, tokens, calls and/or wall-clock seconds.No findings
Why it is in this group
- The name "open_budget" says "open": a word for a change.
Description, as the server wrote it Open a hard budget for one agent run before it starts: max dollars, tokens, calls and/or wall-clock seconds. Returns a handle. Give the handle to every sub-agent; they all draw from the same ledger, so one runaway loop or fork cannot exceed the run cap.
recordAfter a call, report the actual usage minus what you reserved (positive if it cost more, negative to give back an overestimate).1 finding
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
Findings
- High
The input schema at properties.handle.description contains "<secret>", a tag of the kind used to mark instructions for the model.
What this check looks forMatched text
inputSchema.properties.handle.description: "...le from open_budget (looks like bg_<id>.<secret>). Pass it to every sub-agent so they al..."
Description, as the server wrote it After a call, report the actual usage minus what you reserved (positive if it cost more, negative to give back an overestimate). Keeps the ledger honest without blocking.
reserveCall BEFORE each model call, tool call or sub-agent spawn with your estimate.1 finding
Why it is in this group
- The description's first verb is "call": a word for a change.
Findings
- High
The input schema at properties.handle.description contains "<secret>", a tag of the kind used to mark instructions for the model.
What this check looks forMatched text
inputSchema.properties.handle.description: "...le from open_budget (looks like bg_<id>.<secret>). Pass it to every sub-agent so they al..."
Description, as the server wrote it Call BEFORE each model call, tool call or sub-agent spawn with your estimate. Returns allowed=true and deducts it, or allowed=false with the limit that tripped. When allowed is false, stop, do not retry, and return your partial result to the user. Atomic, so parallel sub-agents cannot both slip under the cap.
weekly_reportDraft a plain-English report of the last 7 days for your key (or IP): runs, spend, runs that went over their cap, and dollars spent after a cap would have stopped them.No findings
Why it is in this group
- Nothing in its name or description says what it does, so it is counted as a write.
Description, as the server wrote it Draft a plain-English report of the last 7 days for your key (or IP): runs, spend, runs that went over their cap, and dollars spent after a cap would have stopped them. Read-only; sends nothing. Show it to your user.
Only reads 2Claude Code: allow
budget_statusSee spend vs limits, remaining budget, whether the run is stopped and why, and the last 20 ledger events.1 finding
Why it is in this group
- The name "budget_status" says "status": a word for reading.
Findings
- High
The input schema at properties.handle.description contains "<secret>", a tag of the kind used to mark instructions for the model.
What this check looks forMatched text
inputSchema.properties.handle.description: "...le from open_budget (looks like bg_<id>.<secret>). Pass it to every sub-agent so they al..."
Description, as the server wrote it See spend vs limits, remaining budget, whether the run is stopped and why, and the last 20 ledger events. Set close=true when the run finishes.
usage_and_upgradeShow today's usage for your key or IP and the $49/month All Access upgrade link (cloudforbots.com/offer).No findings
Why it is in this group
- The description's first verb is "show": a word for reading.
Description, as the server wrote it Show today's usage for your key or IP and the $49/month All Access upgrade link (cloudforbots.com/offer). Free tier: 500 calls/day. Not metered.
This is not a safety rating, and no finding is not a safety guarantee. The groups and findings come from fixed checks of the tool definitions the server returned on the date shown, not from its code or what it does when called. How the checks work.
Exfil surface
Which tools read private data, which read text someone else can write, and which can send data out. One of each is a path a prompt injection can use.
No private-data, untrusted-input, outbound path in this server alone: no tool is tagged private-data or untrusted-input or outbound.
- Reads private data0
- Reads files, mail, messages, databases, secrets, calendars or repositories.
- Reads text others write0
- Brings in text someone else can write: web pages, inbound email, messages, issues and comments.
- Can send data out0
- Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL.
Other servers in the same agent add their own tools to these paths. Audit your whole setup to see them together.
Install
Pinned to version 1.0.1. Values in angle brackets are yours to fill in.
claude mcp add --transport http budget https://budget.cloudforbots.com/mcpRun in a terminal.
Claude Code permissions
Allow the reads, ask before a change or a way out, deny the deletes. Paste into .claude/settings.json.
{
"permissions": {
"allow": [
"mcp__budget__budget_status",
"mcp__budget__usage_and_upgrade"
],
"ask": [
"mcp__budget__open_budget",
"mcp__budget__reserve",
"mcp__budget__record",
"mcp__budget__weekly_report"
],
"deny": []
}
}Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. A read that can send data out (it takes a URL, say) asks first too. It assumes the server is installed as budget. Review it before you use it.