Configuration

Agent-Gantry works with zero configuration, but production systems usually configure routing, persistence, execution, telemetry, and provider extras.

AgentGantryConfig

from agent_gantry import AgentGantry
from agent_gantry.schema.config import AgentGantryConfig

config = AgentGantryConfig(
    vector_store={"provider": "lancedb", "path": ".gantry/tools.lance"},
    embedder={"provider": "nomic"},
    routing={"score_threshold": 0.5, "use_llm_for_intent": False},
    execution={"default_timeout_ms": 30000, "max_retries": 2},
    telemetry={"provider": "console"},
)
gantry = AgentGantry(config=config)

Common extras

uv add "agent-gantry[openai]"          # OpenAI, Azure OpenAI, OpenRouter
uv add "agent-gantry[anthropic]"       # Claude tools
uv add "agent-gantry[google-genai]"    # Gemini function calling
uv add "agent-gantry[lancedb,nomic]"   # local durable semantic search
uv add "agent-gantry[mcp,a2a]"         # protocol bridges
uv add "agent-gantry[agent-frameworks]"