Anthropic Agent SDK Integration¶

Engramia provides engramia_query(), recall_system_prompt(), and engramia_hooks() for the Anthropic Agent SDK (claude-agent-sdk).

Installation¶

pip install "engramia[openai,anthropic-agents]"

Quick start¶

from engramia import Memory
from engramia.providers import OpenAIProvider, OpenAIEmbeddings, JSONStorage
from engramia.sdk.anthropic_agents import engramia_query

mem = Memory(
    llm=OpenAIProvider(model="gpt-4.1"),
    embeddings=OpenAIEmbeddings(),
    storage=JSONStorage(path="./engramia_data"),
)

# One-liner: recall → run → learn
async for message in engramia_query(mem, prompt="Build a CSV parser"):
    print(message)

What happens behind the scenes:

Recalled patterns are injected into the system prompt
Agent executes normally, yielding messages
On ResultMessage, the output is stored as a success pattern

engramia_query parameters¶

Parameter	Default	Description
`base_system_prompt`	`""`	Base system prompt to extend with recalled context
`recall_limit`	`3`	Max patterns to recall
`auto_learn`	`True`	Store the final result as a pattern
`min_score`	`7.0`	Eval score assigned when learning
`options`	`None`	Optional `ClaudeAgentOptions` to extend

Manual system prompt injection¶

For full control over the agent options:

from claude_agent_sdk import ClaudeAgentOptions, query
from engramia.sdk.anthropic_agents import recall_system_prompt

prompt = recall_system_prompt(
    mem,
    task="Build a CSV parser",
    base="You are a senior developer.",
    recall_limit=5,
)

async for message in query(
    prompt="Build a CSV parser",
    options=ClaudeAgentOptions(
        system_prompt=prompt,
        allowed_tools=["Read", "Edit", "Bash"],
    ),
):
    print(message)

PostToolUse hooks¶

For observability on tool usage:

from claude_agent_sdk import ClaudeAgentOptions, ClaudeSDKClient
from engramia.sdk.anthropic_agents import engramia_hooks, recall_system_prompt

hooks = engramia_hooks(mem, auto_learn=True)
prompt = recall_system_prompt(mem, task="Build a CSV parser", base="You are a developer.")

options = ClaudeAgentOptions(
    system_prompt=prompt,
    hooks=hooks,
)

async with ClaudeSDKClient(options=options) as client:
    await client.query("Build a CSV parser")
    async for message in client.receive_response():
        print(message)

With custom ClaudeAgentOptions¶

from claude_agent_sdk import ClaudeAgentOptions

options = ClaudeAgentOptions(
    allowed_tools=["Read", "Write", "Bash"],
    max_turns=10,
)

async for message in engramia_query(
    mem,
    prompt="Build a web scraper",
    options=options,
    recall_limit=5,
    min_score=8.0,
):
    print(message)