Agent Spend Guard — owner-set purchasing guardrails
by a2awire.com · remote · no sign-in to list tools · v0.1.0 · published by com.a2awire
Free spend guardrails for AI agents: approve/deny/ask_user, caps, dupes.
4 tools: 2 do something that can't be undone and 2 change things.
Scanned Oct 21, 2026
- Deletes or cancels0Claude Code: deny
- Can't be undone2Claude Code: ask: see these tools
- Changes things2Claude Code: ask: see these tools
- Only reads0Claude Code: allow
No check matched anything in the tool definitions. No finding is not a safety guarantee. How the checks work.
claude mcp add --transport http agent-spend-guard https://a2awire.com/mcp/connectors/spend-guard/httpTools
Open a tool for its reasons and full description. Descriptions are the publisher's own words, shown as they are.
4 tools
Can't be undone 2Claude Code: ask
spend_guard_check_purchaseHumans: sign up with your email at https://a2awire.com/api-key and create an API key, then hand it to your agent (it sends Authorization: Bearer <key> or X-API-Key with every call)…No findings
Why it is in this group
- The name "spend_guard_check_purchase" says "purchase": a word that usually means money moves, which cannot be taken back.
Description, as the server wrote it Humans: sign up with your email at https://a2awire.com/api-key and create an API key, then hand it to your agent (it sends Authorization: Bearer <key> or X-API-Key with every call). Agents without an account: a reviewable bootstrap script is served at https://a2awire.com/api/v1/connectors/spend-guard/bootstrap.sh (download it, read it, then run it to mint a key). Call BEFORE checkout. Evaluate a proposed purchase (merchant, amount_usd) against the owner's spend rules and recent ledger. Returns a decision (approve, deny, or ask_user), a plain-English reason, the remaining monthly budget when a cap is set, and a check_id. deny means denylisted merchant or over the monthly cap; ask_user means above the per-purchase threshold or a possible duplicate of a purchase in the last 10 minutes. Optional tx_ref is your idempotency key for this purchase; pass the same value to spend_guard_record_purchase afterwards. Appends a check event; checks never consume budget. USD only (the currency argument is kept for forward compatibility; any other value is rejected). Not guest-callable. REST: POST /api/v1/spend-guard/check.
Annotations the server declares
- title: Spend Guard Check Purchase
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: false
- openWorldHint: false
spend_guard_record_purchaseHumans: sign up with your email at https://a2awire.com/api-key and create an API key, then hand it to your agent (it sends Authorization: Bearer <key> or X-API-Key with every call)…No findings
Why it is in this group
- The name "spend_guard_record_purchase" says "purchase": a word that usually means money moves, which cannot be taken back.
Description, as the server wrote it Humans: sign up with your email at https://a2awire.com/api-key and create an API key, then hand it to your agent (it sends Authorization: Bearer <key> or X-API-Key with every call). Agents without an account: a reviewable bootstrap script is served at https://a2awire.com/api/v1/connectors/spend-guard/bootstrap.sh (download it, read it, then run it to mint a key). Call AFTER checkout. Record a completed purchase in the owner's spend ledger. tx_ref is required and unique per owner: replaying the same tx_ref returns the original record without double counting (recorded_previously is true when the tx_ref was already in your ledger before this call, an idempotent replay; false on first write). Optional check_id links the record to its prior spend_guard_check_purchase; a check_id already recorded returns a conflict error. Returns the recorded event and the month-to-date total. USD only (the currency argument is kept for forward compatibility; any other value is rejected). Not guest-callable. REST: POST /api/v1/spend-guard/record.
Annotations the server declares
- title: Spend Guard Record Purchase
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
Changes things 2Claude Code: ask
spend_guard_reportHumans: sign up with your email at https://a2awire.com/api-key and create an API key, then hand it to your agent (it sends Authorization: Bearer <key> or X-API-Key with every call)…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.
- The server marks it read-only (readOnlyHint: true), but an annotation never lowers a class.
Description, as the server wrote it Humans: sign up with your email at https://a2awire.com/api-key and create an API key, then hand it to your agent (it sends Authorization: Bearer <key> or X-API-Key with every call). Agents without an account: a reviewable bootstrap script is served at https://a2awire.com/api/v1/connectors/spend-guard/bootstrap.sh (download it, read it, then run it to mint a key). Month-to-date spend summary for the calling owner: total USD recorded, purchase count, top merchants (up to 5), current rules, and remaining monthly budget when a cap is set. Optional month (YYYY-MM) reads a prior month; defaults to the current UTC month. Not guest-callable. REST: GET /api/v1/spend-guard/report.
Annotations the server declares
- title: Spend Guard Report
- readOnlyHint: true
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
spend_guard_set_rulesHumans: sign up with your email at https://a2awire.com/api-key and create an API key, then hand it to your agent (it sends Authorization: Bearer <key> or X-API-Key with every call)…No findings
Why it is in this group
- The name "spend_guard_set_rules" says "set": a word for a change.
Description, as the server wrote it Humans: sign up with your email at https://a2awire.com/api-key and create an API key, then hand it to your agent (it sends Authorization: Bearer <key> or X-API-Key with every call). Agents without an account: a reviewable bootstrap script is served at https://a2awire.com/api/v1/connectors/spend-guard/bootstrap.sh (download it, read it, then run it to mint a key). Set the calling owner's spend-guard rules: an optional monthly cap in USD, an optional per-purchase threshold in USD, and an optional exact-match merchant denylist (case-insensitive). Full declaration: omit or null a field to clear that rule. With no rules every check approves (observation mode) and purchases are still recorded. Returns the effective rules. Not guest-callable. REST: PUT /api/v1/spend-guard/rules.
Annotations the server declares
- title: Spend Guard Set Rules
- readOnlyHint: false
- destructiveHint: false
- idempotentHint: true
- openWorldHint: false
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.
Install
Pinned to version 0.1.0. Values in angle brackets are yours to fill in.
claude mcp add --transport http agent-spend-guard https://a2awire.com/mcp/connectors/spend-guard/httpRun in a terminal.
Claude Code permissions
Allow the reads, ask before a change, deny the deletes. Paste into .claude/settings.json.
{
"permissions": {
"allow": [],
"ask": [
"mcp__agent-spend-guard__spend_guard_check_purchase",
"mcp__agent-spend-guard__spend_guard_record_purchase",
"mcp__agent-spend-guard__spend_guard_set_rules",
"mcp__agent-spend-guard__spend_guard_report"
],
"deny": []
}
}Tools that only read are allowed, tools that change things ask first, tools that delete or cancel are denied. It assumes the server is installed as agent-spend-guard. Review it before you use it.