Follow the whole system.
Understand direct agents, composite agents, and their communication paths. Keep each agent’s health distinct from the supporting services it depends on.
Read the health contracts → OPEN SOURCE. AGENT NATIVE.Running is easy.
Running is easy.
Ready is different.
Know your agents can do the work. A universal health and readiness framework for AI agents, the systems they talk to, and everything they depend on.
Vendor neutral / Framework neutral / Apache 2.0
$ agenthealth check production.yaml
DEPENDENCY GRAPH ILLUSTRATIVE OUTPUT
◆ Direct Agent HEALTHY
├── Composite Agents HEALTHY
├── MCP Servers HEALTHY
├── Generative AI Models HEALTHY
└── Vector Databases HEALTHY
One system. Shared health signals.
ONE HEALTH LANGUAGE.
YOUR ENTIRE STACK.
YOUR ENTIRE STACK.
Agents
MCP
A2A
Models
Data
Kubernetes
BEYOND A 200 OK
Health checks that understand
how agents actually work.
A running process is only the beginning. Connect agent readiness to capability, protocol, and dependency health with consistent, actionable evidence.
Speak the same language.
Bring HTTP APIs, MCP servers, A2A endpoints, gateways, and routers into one health model. Use the Agent Health Protocol to share readiness signals.
Read the health contracts →See what readiness relies on.
Trace models, tools, databases, and vector stores through dependency graphs. Turn a collection of individual checks into a meaningful system view.
Read the health contracts →Find the failing layer.
Separate reachability, protocol, configuration, and capability evidence so the next action is clear.
Explore doctor →Trace dependency impact.
See how shared dependencies and independent edge policies shape the readiness of a cooperating system.
Understand the graph →Inspect the connections.
Distinguish gateway signals, router backends, and configured communication paths from agent health.
Explore routing checks →Make health actionable.
Use the same checks locally and in deployment automation, with structured results and predictable exit codes.
Use the CLI →Expose readiness signals.
Serve health and readiness through the Agent Health Protocol, with aggregate signals and authenticated detail.
Explore AHP →Build a shared foundation.
Extend a vendor-neutral framework with adapter contracts and common health semantics for your ecosystem.
Read the specification →HEALTH ACROSS THE AGENTIC SYSTEM
Every connection matters.
Check them from the center.
AgentHealth checks agents and supporting dependencies across your system, collecting the evidence behind one readiness decision.
User & ApplicationPrompt · query · response
HEALTH & READINESS
AgentHealth
Ping. Diagnose. Understand.
14 / 14 checks healthyOne readiness decisionIndependent evidence for every layer.
MODEL DEPENDENCIES✓
Generative AI Models
LLMs · inference · model availability
AGENT HEALTH✓
Direct Agent
First agent · liveness · readiness
AGENT STATE & REASONING✓
Memory & Planning
Short-term memory · long-term memory · planning
MULTI-AGENT SYSTEM✓
Composite Agents
Downstream agents · capabilities · task probes
GATEWAY HEALTH✓
Agentgateway
Gateway signals · backends · protocol paths
AGENT-TO-AGENT PROTOCOL✓
A2A Endpoints
Agent discovery · skills · communication
ROUTING HEALTH✓
Agent Router
Router signals · named routes · backends
MODEL CONTEXT PROTOCOL✓
MCP Servers
Initialization · tools · resources · prompts
RETRIEVAL DEPENDENCIES✓
Vector Databases
Connectivity · collections · retrieval
FUNCTIONAL DEPENDENCIES✓
Tools
Tool discovery · availability · safe invocation
DATA DEPENDENCIES✓
Databases
Connectivity · authentication · read readiness
SEARCH DEPENDENCIES✓
Search Engines
Search APIs · index availability · latency
COMPUTE & ORCHESTRATION✓
Agent Runtimes
Frameworks · containers · cloud infrastructure
SERVICE DEPENDENCIES✓
HTTP APIs
DNS · TLS · authentication · HTTP status
A SHARED MODEL
Ask better questions.
Get useful answers.
From a laptop to a production cluster, use consistent checks and machine-readable results to decide what happens next.
Explore the specification →01
Can I reach it?
Endpoint and transport reachability.
02
Can we communicate?
Protocol negotiation and compatibility.
03
Is it ready to work?
Configuration, capabilities, and dependencies.
04
Can it do the job?
Explicit, opt-in functional probes.
BUILT TO FIT YOUR WORKFLOW
Local checks. Deployment gates.
Production readiness.
CLI & containers
A portable health engine for terminals, automation, and containerized workloads.
Python & TypeScript
Bring the shared health model into applications and agent frameworks.
Kubernetes & CI/CD
Make readiness part of probes, deployment validation, and promotion decisions.
Observability & extensions
Connect health evidence to telemetry, adapters, and ecosystem conformance.
YOUR FIRST HEALTH CHECK
From download
to diagnosis.
Download the binary for your platform, verify its checksum, and check an HTTP endpoint you control. No language runtime required.
Installation & verification guide →TERMINAL
# Check your installation
./agenthealth version
# Check your local API
./agenthealth ping http http://localhost:8080/health
# Run configured checks and inspect results
./agenthealth check agenthealth.yaml --format json
./agenthealth doctor agenthealth.yamlCreate agenthealth.yaml before running configured checks.
Thoughtful defaults. Explicit actions.
Passive checks first. Active probes opt in. Bounded timeouts, TLS verification, and credential redaction built into the health engine.
PICK YOUR STARTING POINT
Less searching.
More building.
01 / GUIDE
Run your first check
Install a binary and learn the core commands.
02 / GUIDEConnect MCP servers
Inspect protocol endpoints and capabilities.
03 / GUIDEModel your dependencies
Define relationships and readiness policies.
04 / GUIDETry a working example
Start with sample configurations you can adapt.
OPEN SOURCE, SHARED FOUNDATION
Better agent health
starts with all of us.
Build an adapter. Improve the specification. Share an integration.
Help make readiness a common language for the agentic ecosystem.