Is this MCP server or tool safe to call right now? About 2,000 servers, probed every 15 minutes.
check_mcp_server_health
find_alternative_mcp_servers
watch_server
list_watches
unwatch_server
get_api_key
Upstream
API status and model retirements, across about 40 providers.
check_api_status
check_model_deprecation
list_recent_api_changes
list_providers
get_digest
SpendGate
approve, ask_human, or block before an agent pays.
score_spend
get_quota
spendgate_get_api_key
Copy this MCP config
Free to try. No signup. These configs connect with no key (50 calls/day) and no Authorization header. Call get_api_key to raise the quota to 1,000 calls/day and 3 watches, then send that key as X-API-Key.
How often a verdict still held at the very next scheduled check of the same server (usually ~15 minutes later). This measures how stable our verdicts are between checks, not whether real tool calls succeed. History starts Oct 15, 2026, so samples are still small. Raw data
Loading…
Get a free API key
No signup. With no key you get 50 checks/day. get_api_key (or this button, or POST /v1/keys) raises that to 1,000 checks/day and 3 change-alert watches. The key is shown once.
Pricing
Zero signup. Agents start free (no key needed), then pay per call in USDC on Base via x402 — no account, no subscription. You're never charged for errors.
Every metered response carries X-Free-Remaining and X-Price-After-Free (and a quota body field) so agents can plan. Machine-readable: /v1/pricing · /.well-known/x402
For agents: ask before you call
curl -H "X-API-Key: $TW_KEY" \
"https://toolweather.toolweather.workers.dev/v1/is_safe_to_call?server=io.github.toolweather/spendgate"
{ "verdict": "safe", "status": "ok", "latency_ms": 182,
"spec_readiness": {"readiness": "dual", "supports_2026_07_28": true},
"last_tool_change": null, "last_checked": "…" }
# or add the MCP server and call the tools:
check_mcp_server_health(server="io.github.toolweather/spendgate", tool="search")
check_api_status(provider="openai")