Stop AI agents from
burning your budget

AgentKavach wraps your LLM client with hard budget enforcement, real-time alerts, and a kill switch that fires before you overspend. One SDK for OpenAI, Anthropic, Google, and Mistral.

Start Free Documentation

< 0.1ms

Budget check overhead

LLM providers

Alert channels

99.9%

Uptime SLA

from agentkavach import AgentKavach, Budget, ChannelType

def emergency_stop():
    agent.save_checkpoint()
    sys.exit(1)

guard = AgentKavach(
    api_key="cg_...",
    llm_key="sk-...",
    budget=Budget.daily(50),
    on_kill=emergency_stop,
)

response = guard.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello!"}],
)

How it works

Three steps to hard budget enforcement. No infrastructure changes required.

STEP 01

Install the SDK

One pip install. No infrastructure changes, no sidecar processes.

pip install agentkavach

STEP 02

Wrap your LLM client

Replace your provider client with AgentKavach. Same API, same parameters, with budget enforcement built in.

guard = AgentKavach(
    api_key="cg_...",
    llm_key="sk-...",
    budget=Budget.daily(50),
    on_kill=emergency_stop,
)

STEP 03

Ship with confidence

Every call is tracked. Alerts fire at your thresholds. The kill switch engages at 100%. No surprises on your bill.

response = guard.create(
    model="gpt-4o",
    messages=[...],
)

Everything you need to control AI costs

From a single developer to a fleet of 10,000 agents. AgentKavach scales with your team.

Hard Budget Limits

Set daily, monthly, or total budgets per agent. The kill callback fires at 100% to halt runaway spend before it escalates.

Multi-Provider

OpenAI, Anthropic, Google, Mistral. Track costs across all providers with a single SDK and a unified guard.create() API.

Real-Time Alerts

Slack, email, PagerDuty, or custom webhooks. Configure escalating thresholds at 50%, 80%, 95%, and 100%.

Team Dashboards

See every agent's spend in real time. Drill into model usage, cost trends, and alert history from a single pane.

Guardrails

Token limits, call count caps, runtime limits, and runaway loop detection. Stop agents before they spiral out of control.

Shared Budgets

Pool a single budget across multiple agents and providers. Enforce team-wide spend limits with thread-safe accounting.

One SDK, every provider

The same guard.create() call works across OpenAI, Anthropic, Google, and Mistral. Switch providers without changing your budget logic.

OpenAI

Anthropic

Google

Mistral

from agentkavach import AgentKavach, Budget

guard = AgentKavach(
    provider="openai",       # "openai", "anthropic", "google", or "mistral"
    api_key="cg_...",        # your AgentKavach API key
    llm_key="sk-...",        # your LLM provider API key
    agent_name="my-agent",   # identifies this agent in the dashboard
    budget=Budget.daily(50), # $50/day hard limit
    on_kill=emergency_stop,  # callback at 100% utilization
    save_prompts=False,      # opt-in: log prompts to dashboard
)

response = guard.create(
    model="gpt-4o",          # any model from any supported provider
    messages=[{"role": "user", "content": "Hello!"}],
)

Escalating alerts, automatic shutdown

Configure alert channels at different budget thresholds. Start with an email at 50%, escalate to Slack at 80%, page your on-call at 95%, and trigger the kill switch at 100%. Each channel validates its credentials at construction time.

50%

Email notification to ops team

80%

Slack message to #cost-alerts channel

95%

PagerDuty incident for on-call team

100%

Kill switch fires, agent halted

guard = AgentKavach(
    provider="openai",
    api_key="cg_...",
    llm_key="sk-...",
    budget=Budget.daily(100),
    on_kill=emergency_stop,
    channels=[
        AgentKavach.channel(ChannelType.EMAIL,
            threshold=0.50,
            to="ops@acme.com",
        ),
        AgentKavach.channel(ChannelType.SLACK,
            threshold=0.80,
            webhook_url="https://hooks...",
            template="{agent_name} at {pct}% budget",
        ),
        AgentKavach.channel(ChannelType.PAGERDUTY,
            threshold=0.95,
            routing_key="R0abc...",
        ),
        AgentKavach.channel(ChannelType.KILL,
            threshold=1.0,
        ),
    ],
)

agentkavach.yaml

channels:
  slack:
    type: slack
    webhook_url: https://hooks.slack.com/...
    channel: "#cost-alerts"
  pagerduty:
    type: pagerduty
    routing_key: R0abc123
    service_url: https://events.pagerduty.com/v2/enqueue
    team: AI Ops Team

shared_budgets:
  engineering-pool:
    limit: 500
    period: daily
    agents: [research-bot, code-review]

agents:
  research-bot:
    provider: openai
    budget: { type: daily, limit: 50 }
    alerts:
      - { threshold: 0.80, channels: [slack] }
      - { threshold: 1.0, channels: [kill] }

  support-agent:
    provider: anthropic
    budget: default

Manage entire teams with YAML

Define budgets, alerts, and shared pools for all your agents in a single configuration file. AgentKavach validates every channel reference, budget type, and shared pool name at load time. Misconfigured alerts fail fast, not at 3 AM.

Shared budgets across agents and providers

Channel definitions with type validation

Defaults with per-agent overrides

Load all agents with one AgentKavach.from_yaml() call

Security

AgentKavach is built with security at every layer. Your API keys, spend data, and agent communications are protected by industry-standard practices from day one. We never use your LLM provider API key for any purpose other than proxying your requests. We only read the response metadata (token counts, model used) to track costs — we never store, log, or access your prompts or completions.

TLS 1.2+ on all API traffic
JWT + API key authentication
Role-based authorization, keys scoped per org
Parameterized queries, no SQL injection
Redis-backed rate limiting
HMAC webhook signing
Your provider API key is never used for any purpose beyond proxying your requests

Simple, transparent pricing

Start free. Scale when you need to.

Free

$0forever

For solo developers and prototypes.

Up to 3 agents
10,000 events/day
Email alerts
Kill switch

Pro

$20/month

For teams running production agents.

Up to 100 agents
1 million events/day
Email alerts
Kill switch
Agent-level budgets

Start Trial

Max

$100/month

For fleets of AI agents at scale.

Up to 10,000 agents
1,000,000 events/day
All alert channels (Slack, PagerDuty, Webhooks)
Kill switch
Shared budgets across teams

Contact Sales

Model Pricing Reference

Sourced from provider pricing pages: OpenAI, Anthropic, Google, Mistral

Provider	Model	Input / 1K tokens	Output / 1K tokens
Anthropic	claude-3-5-haiku-20241022	$0.000800	$0.004000
Anthropic	claude-3-5-sonnet-20241022	$0.003000	$0.015000
Anthropic	claude-3-haiku-20240307	$0.000250	$0.001250
Anthropic	claude-3-opus-20240229	$0.015000	$0.075000
Anthropic	claude-haiku-4-5	$0.000800	$0.004000
Anthropic	claude-opus-4-0	$0.015000	$0.075000
Anthropic	claude-opus-4-6	$0.015000	$0.075000
Anthropic	claude-sonnet-4-0	$0.003000	$0.015000
Anthropic	claude-sonnet-4-6	$0.003000	$0.015000
Google	gemini-1.5-flash	$0.000075	$0.000300
Google	gemini-1.5-pro	$0.001250	$0.005000
Google	gemini-2.0-flash	$0.000100	$0.000400
Google	gemini-2.5-flash	$0.000150	$0.003500
Google	gemini-2.5-pro	$0.001250	$0.010000
Mistral	codestral-latest	$0.000300	$0.000900
Mistral	ministral-8b-latest	$0.000100	$0.000100
Mistral	mistral-embed	$0.000100	$0.000100
Mistral	mistral-large-latest	$0.002000	$0.006000
Mistral	mistral-small-latest	$0.000100	$0.000300
Mistral	pixtral-large-latest	$0.002000	$0.006000
OpenAI	codex-mini	$0.001500	$0.006000
OpenAI	gpt-3.5-turbo	$0.000500	$0.001500
OpenAI	gpt-4	$0.030000	$0.060000
OpenAI	gpt-4-turbo	$0.010000	$0.030000
OpenAI	gpt-4.1	$0.002000	$0.008000
OpenAI	gpt-4.1-mini	$0.000400	$0.001600
OpenAI	gpt-4.1-nano	$0.000100	$0.000400
OpenAI	gpt-4.5-preview	$0.075000	$0.150000
OpenAI	gpt-4o	$0.002500	$0.010000
OpenAI	gpt-4o-mini	$0.000150	$0.000600
OpenAI	o1	$0.015000	$0.060000
OpenAI	o1-mini	$0.003000	$0.012000
OpenAI	o3	$0.010000	$0.040000
OpenAI	o3-mini	$0.001100	$0.004400
OpenAI	o4-mini	$0.001100	$0.004400

Frequently Asked Questions (FAQ)

Can I change plans anytime?

Yes. Upgrade or downgrade at any time. Changes take effect immediately. Downgrades are prorated.

What counts as an event?

Each LLM API call tracked through the SDK counts as one event. Batch calls count as one event per item in the batch.

Is there a free trial for paid plans?

Pro comes with a 14-day free trial. No credit card required to start. Max does not include a trial.

What happens when I hit my event limit?

Once you exceed your event limit, monitoring, alerts, and the kill switch stop functioning until your limit resets at midnight. New agents cannot be onboarded past the limit either. To avoid disruption, upgrade to a higher plan before reaching your cap.

Do you support custom invoicing?

Max customers can request custom invoicing and payment terms. Contact us for details.

Start controlling costs in 2 minutes

pip install agentkavach

Create Your Account

Stop AI agents fromburning your budget