diff --git a/site/index.html b/site/index.html index 801a085..c90bd24 100644 --- a/site/index.html +++ b/site/index.html @@ -3,22 +3,22 @@ - AgentGuard — AI Agent Cost Control | Budget Limits & Spend Visibility - + AgentGuard — Runtime Firewall for AI Agents + - - + + - - + + @@ -195,20 +195,28 @@
- Open Source · Cost Control for AI Agents -

Stop runaway AI bills before they happen.

-

Set hard spend limits, catch looping agents, and see cost per run. Open-source SDK + hosted dashboard.

+ Runtime Firewall for AI Agents +

Your AI agent just burned $2,000 in tokens overnight.
AgentGuard stops it at $5.

+

Stop runaway AI agents before they burn your API budget or leak data.

+
- Get Started — it's free + Start Your 14-Day Free Trial View on GitHub
- Works with + Built for teams running AI agents in production: LangChain - OpenAI - Anthropic + LangGraph + CrewAI + AutoGen + MCP Agents Any Python agent
@@ -233,23 +241,29 @@

Stop runaway AI bills before they happen.

- Hard budget caps -

Set a dollar limit per project or agent. Runs stop when the cap is hit. No surprise invoices.

+ Kill infinite loops +

Detect repeated tool calls across any depth. Automatic circuit breaker fires before costs compound.

- Loop & runaway detection -

Catch repeated tool calls and stuck agents automatically. Circuit breaker built in.

+ Enforce hard dollar caps +

Set a budget. Agent stops mid-run when it's hit. No more $2,000 surprise invoices.

- Spend visibility -

See cost per agent run, per model, per day. Know where every dollar goes.

+ Track cost per agent +

See exactly what each run costs, per LLM call. Dollar amounts, not token estimates.

- Policy management -

Configure budget caps, loop limits, and timeouts from the dashboard. Push enforcement to your SDK.

+ Remote kill switch +

Stop any agent instantly from the dashboard. No redeployment needed.

+ +
+

Docker gave you container scanning.
AgentGuard gives you agent scanning.

+

CI/CD gate → Runtime enforcement → Compliance trail

+
+

How it works

@@ -290,6 +304,82 @@

How it works

pass # your tool here
+ +

See it in action

+
+
+ +
+ + + +

Other tools watch your agents fail. AgentGuard stops them.

+
+ + + + + + + + + + + + + + + + + + + + +
FeatureAgentGuardLangfuseLangSmithHelicone
Tracing
Cost tracking
Budget enforcementHard stop mid-runSoft alert
Loop detectionAuto circuit-breaker
Remote kill switchInstant, no redeploy
Auto-interventionRule engine on ingest
Open sourceSDK (MIT)Full (MIT)Gateway
Setup3 linesHours30 min2 min
+
+ + +
+

"Can't I just add max_iterations myself?"

+

You could add a budget check. But can you also detect retry loops across 12 tool calls, auto-kill the agent remotely without redeploying, send an alert with the trace link, log the intervention for compliance, and show your team a dashboard of how much you saved — all in a weekend? That's what AgentGuard does out of the box.

+
+

What builders say

@@ -356,7 +446,7 @@

Security & Trust

Retention you control -

7-90 day auto-cleanup by plan. No indefinite data hoarding.

+

30-90 day auto-cleanup by plan. No indefinite data hoarding.