Documentation

Everything you need to integrate AgentKavach into your AI agent stack.

Quickstart

Install the SDK and make your first budget-protected LLM call in under 2 minutes.

Full examples for OpenAI, Anthropic, Google, and Mistral with model lists and cross-provider comparison.

Daily, monthly, and total budget types with in-memory enforcement and fail-open design.

Pool a single budget across multiple agents and providers with thread-safe enforcement.

Token limits, call count caps, runtime limits, and loop detection to keep agents in check.

Slack, email, PagerDuty, and webhook alerts with configurable thresholds and templates.

Define a fleet of agents, shared pools, and channels in a single file.

REST endpoints for event ingestion, dashboard data, agent management, and billing.

Complete list of SDK exceptions and API error codes with troubleshooting guidance.

Per-token pricing tables for every supported model across all providers.

OpenAI and Anthropic streaming with chunk-level token counting and GeneratorExit handling.

Current AgentKavach release. Full history lives in the GitHub repository.