getprimer 0.1.0

Persistent memory and governance layer for AI agents

getprimer

Persistent memory and governance layer for AI agents.

pip install getprimer

Quick start

from getprimer import Primer

p = Primer(api_key="pk_live_...")  # or set PRIMER_API_KEY env var

# ── Always do this pattern ─────────────────────────────────────────
ctx = p.brief(agent="my-bot")           # load context at session start
print(ctx["context_summary"])           # inject into system prompt

# ... agent does work ...

p.learn("Customer prefers async comms", pinned=True, source_trust="human")
p.check("delete_records")              # True = allowed, False = blocked
p.handoff("Done. Next: add tests.")    # save handoff at session end
# ──────────────────────────────────────────────────────────────────

LangChain

from getprimer.langchain import PrimerTools

tools = PrimerTools(api_key="pk_live_...").as_tools()
agent = initialize_agent(tools=tools, llm=ChatOpenAI(), ...)

Autonomous mode (zero explicit calls — Primer hooks in automatically):

from getprimer.langchain import PrimerCallbackHandler

callback = PrimerCallbackHandler(api_key="pk_live_...")
llm = ChatOpenAI(callbacks=[callback])
# Every LLM chain now gets Primer context injected automatically

Session limit

On the free plan (20 sessions/month), when the limit is reached brief() raises SessionLimitError. The exception message is safe to surface directly to the user:

from getprimer import Primer, SessionLimitError

try:
    ctx = p.brief()
except SessionLimitError as e:
    print(e)  # "⚠️ Primer: You've used all 20 free sessions..."
              # Upgrade at: https://getprimer.cloud/pricing

Get an API key

Sign up at getprimer.cloud — free plan included.