forgein

Your project memory, available everywhere.
Write your context once in Claude Code. Use it in any Python agent, LLM call, or AI pipeline.

---

The problem

Your Claude Code memory lives in CLAUDE.md. Your Cursor rules in .cursorrules. Your Copilot instructions in .github/copilot-instructions.md. Every AI tool has its own silo — and none of them talk to each other.

When you write a prompt, spin up an agent, or call an API, you start from scratch — re-explaining your stack, your conventions, your current sprint — every single time.

The fix

pip install forgein

from forgein import ForgeinClient

client = ForgeinClient()  # reads FORGEIN_API_TOKEN
context = client.adapters.get("copilot", project="my-saas-app")
# → "# Project Context\n**Stack:** Next.js 15, TypeScript, PostgreSQL..."

One call. One source of truth. Works with Claude, OpenAI, Gemini, LangChain, and anything else that takes a string.

---

Install

# Core SDK (sync + async)
pip install forgein

# With LangChain tool support
pip install "forgein[langchain]"

# With CLI
pip install "forgein[cli]"

# Everything
pip install "forgein[all]"

---

Quick start

1. Get a token

# In Claude Code
/forgein auth

Or grab one at app.forgein.ai/tokens.

2. Set the environment variable

export FORGEIN_API_TOKEN="fg_..."

3. Fetch context

from forgein import ForgeinClient

with ForgeinClient() as client:
    # Get context for any AI tool
    copilot    = client.adapters.get("copilot",     project="my-saas-app")
    cursor     = client.adapters.get("cursor",      project="my-saas-app")
    claude_md  = client.adapters.get("claude-code", project="my-saas-app")

    # Or sync all context files to disk in one call
    result = client.adapters.sync("my-saas-app", path=".")
    # Writes: .github/copilot-instructions.md, .cursorrules,
    #         .windsurfrules, CLAUDE.md, .gemini/context.md
    print(f"{len(result.written)} files written")

---

Integrations

Anthropic / Claude

import anthropic
from forgein import ForgeinClient

context = ForgeinClient().adapters.get("claude-code", project="my-saas-app")

client = anthropic.Anthropic()
message = client.messages.create(
    model="claude-opus-4-8",
    system=context,
    messages=[{"role": "user", "content": "Refactor the auth middleware"}],
    max_tokens=8192,
)
print(message.content[0].text)

Or use the integration helper:

from forgein.integrations.anthropic import get_system_prompt

system = get_system_prompt("my-saas-app")  # defaults to claude-code adapter

OpenAI

from openai import OpenAI
from forgein.integrations.openai import get_system_message

client = OpenAI()
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[
        get_system_message("my-saas-app"),  # → {"role": "system", "content": "..."}
        {"role": "user", "content": "How should I structure the Stripe webhook?"},
    ],
)

LangChain

from langchain_anthropic import ChatAnthropic
from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain_core.prompts import ChatPromptTemplate
from forgein.integrations.langchain import ForgeinContextTool

tools = [ForgeinContextTool(project="my-saas-app")]

llm = ChatAnthropic(model="claude-opus-4-8")
prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful coding assistant."),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}"),
])

agent = create_tool_calling_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools)
result = executor.invoke({"input": "What are the Drizzle ORM conventions in this project?"})

Async

import asyncio
from forgein import AsyncForgeinClient

async def main():
    async with AsyncForgeinClient() as client:
        # Fetch three adapters concurrently
        copilot, cursor, claude = await asyncio.gather(
            client.adapters.get("copilot",     project="my-saas-app"),
            client.adapters.get("cursor",      project="my-saas-app"),
            client.adapters.get("claude-code", project="my-saas-app"),
        )

asyncio.run(main())

---

API reference

ForgeinClient(token=None, *, base_url=None, timeout=30.0, max_retries=3)

Parameter	Type	Description
token	str | None	fg_ API token. Falls back to FORGEIN_API_TOKEN.
base_url	str | None	Override API URL. Falls back to FORGEIN_BASE_URL.
timeout	float	Request timeout in seconds. Default: 30.
max_retries	int	Retry attempts on 429 / 5xx. Default: 3.

Supports context manager (with ForgeinClient() as client:).

AsyncForgeinClient has an identical signature and supports async with.

---

client.adapters

Method	Returns	Description
.get(adapter, *, project)	str	Fetch context string for one adapter
.sync(project, *, path=".", adapters=None)	SyncResult	Write context files to disk
.usage()	AdapterUsage	Per-adapter usage stats + 14-day sparkline

Available adapters: "copilot" · "claude-code" · "cursor" · "windsurf" · "gemini" · "chatgpt"

Files written by .sync():

Adapter	File
copilot	.github/copilot-instructions.md
cursor	.cursorrules
windsurf	.windsurfrules
claude-code	CLAUDE.md
gemini	.gemini/context.md
chatgpt	.chatgpt/context.json

---

client.memory

Method	Returns	Description
.list(project)	list[MemoryFileMeta]	List files (no content)
.get(project, path)	MemoryFile	Read a single file
.put(project, path, content)	MemoryFileMeta	Create or update a file
.delete(project, path)	None	Delete a file
.projects()	list[MemoryProject]	All projects with file counts
.search(query)	list[SearchResult]	Full-text search
.stats()	MemoryStats	Aggregate counts and byte sizes
.health()	MemoryHealth	Stale-context detection
.history(project, path)	list[MemoryHistoryEntry]	Version history (max 50)

---

client.tokens

Method	Returns	Description
.list()	list[Token]	All active tokens
.create(name)	TokenCreated	Create a token (raw value returned once)
.revoke(token_id)	None	Revoke by ID

---

client.auth

Method	Returns	Description
.me()	User	Current user
.change_password(current, new)	None	Change password
.export_data()	bytes	Full JSON export
.delete_account()	None	Permanently delete account

---

client.webhooks

Method	Returns	Description
.list()	list[Webhook]	All webhooks
.create(url, events)	Webhook	Create endpoint
.delete(id)	None	Delete
.update(id, *, url, is_active)	Webhook	Update
.deliveries(id)	list[WebhookDelivery]	Last 20 deliveries
.ping(id)	WebhookDelivery	Send test delivery

---

Additional resources

client.contexts · client.shares · client.referrals · client.skills · client.waitlist

All follow the same sync/async pattern. Full reference at forgein.ai/docs/python.

---

Error handling

from forgein import ForgeinClient
from forgein.exceptions import (
    AuthenticationError,
    PermissionDeniedError,
    NotFoundError,
    RateLimitError,
    ServerError,
    NetworkError,
)

client = ForgeinClient()

try:
    context = client.adapters.get("copilot", project="my-saas-app")
except AuthenticationError:
    # Token missing, invalid, or revoked
    print("Run: forgein auth")
except PermissionDeniedError:
    # Pro-only feature
    print("Upgrade at app.forgein.ai/billing")
except RateLimitError as e:
    import time
    time.sleep(e.retry_after or 60)
except NotFoundError:
    print("Project not found — check the name")
except (ServerError, NetworkError) as e:
    print(f"Transient error: {e}")

All exceptions inherit from forgein.exceptions.ForgeinError and expose .status_code and .response.

---

CLI

# Verify auth
forgein auth

# Print context to stdout
forgein context get my-saas-app --adapter copilot

# Sync all context files to current directory
forgein context sync my-saas-app

# List memory files
forgein memory list my-saas-app

# Read a file
forgein memory get my-saas-app CLAUDE.md

# Check API health
forgein health

---

Environment variables

Variable	Description
FORGEIN_API_TOKEN	API token (fg_...)
FORGEIN_BASE_URL	Override API base URL (default: https://api.forgein.ai)

---

Requirements

• Python ≥ 3.9
• httpx >= 0.27
• pydantic >= 2.0

---

License

MIT — see LICENSE.

---

Made by forgein · Get a token · Report an issue