repld-tool 0.0.4

Persistent Python runtime with MCP channel push. Dev shell and autonomous-agent substrate in one package.

repld

Persistent Python runtime with MCP channel push. Dev shell and autonomous-agent substrate in one package.

uv tool install repld-tool

What it does

• Stateful kernel — auth once, hold the client, query across turns. State persists across cells.
• Async-native — top-level await, defer() for fire-and-forget, @every() for periodic tasks. Long jobs never block the turn.
• Channel push — task completion, webhooks, file changes, and timers arrive as <channel> injections. The agent reacts; it doesn't poll.
• Shared namespace — human and agent operate on the same __main__. Stage data in one, use it in the other.
• Browser integration — attach to your logged-in Chrome tabs via CDP. No API keys, no OAuth dance. The agent discovers the API surface from your traffic.
• Gists — reusable Python modules that wrap any web app's API. The browser supplies auth; the gist captures the pattern.

Install

# install globally
uv tool install repld-tool

# in any project:
cd path/to/project
repld init               # writes .mcp.json + updates .gitignore
repld                    # starts the kernel

Project-local alternative: uv add --dev repld-tool, then point .mcp.json at uv run repld bridge.

repld init produces this .mcp.json:

{
  "mcpServers": {
    "repld": { "type": "stdio", "command": "repld", "args": ["bridge"] }
  }
}

Quick example

The agent calls exec to run Python in the kernel:

# runs inline — result returned immediately
import httpx
httpx.get("https://api.example.com/status").json()

# long-running — returns task_id, pushes channel notification on completion
await asyncio.sleep(30)
notify("done", kind="migration")

Autonomous worker — five lines:

@every(300)
async def check_overdue():
    for inv in await po.get_overdue():
        notify(f"Overdue: {inv.customer} {inv.amount} NOK",
               kind="overdue", invoice_id=inv.id)

The kernel runs the watcher; the agent reacts to each <channel> injection.

With an existing app

repld inherits your project's environment. A repl.py at the project root:

from myapp.main import app
from myapp.db import async_session_maker
import asyncio, uvicorn

asyncio.create_task(uvicorn.Server(
    uvicorn.Config(app, host="127.0.0.1", port=8000, log_level="warning")
).serve())

session = async_session_maker()
print("FastAPI on :8000, db session ready")

repld --init repl.py

The agent now has a live handle on your running app: inspect routes, query the ORM, call handlers bypassing HTTP.

Tools

Core:

Tool	What it does
exec	Execute Python. Returns inline within timeout (default 2s); otherwise returns task_id and pushes channel on completion.
get_task	Status + head/tail preview of a running task's output.
cancel	Cancel a running task by id.

Browser (requires uv tool install repld-tool[browser]):

Tool	What it does
browser_attach	Watch URL pattern, auto-attach matching tabs.
browser_tabs	List attached tabs.
browser_pages	List all Chrome targets.
browser_js	Evaluate JavaScript in a tab.
browser_network	Query captured traffic (HAR-style, DuckDB).
browser_body	Response body for a captured request.
browser_request	Request headers/postData for a captured request.
browser_fetch	In-page fetch (inherits auth/cookies).
browser_click	Click element (trusted dispatch).
browser_type	Type into element (trusted dispatch).
browser_key	Send key press (Enter, Escape, etc).
browser_navigate	Navigate tab to URL.
browser_open	Open new tab.
browser_console	Query console logs and exceptions.
browser_screenshot	Capture page screenshot.
browser_cdp	Raw CDP passthrough.
browser_clear	Reset captured network/console.
browser_detach	Remove watch pattern, detach tabs.

Output from every cell spills to $XDG_RUNTIME_DIR/repld/ — the inline response carries a head/tail preview plus the spill path. Use standard Read/Grep tools for full output.

Helpers

Available in the kernel namespace:

notify(content, **meta)              # channel push to the agent
ask(prompt)                          # block on free-form human input
confirm(prompt)                      # block on yes/no
choose(prompt, options)              # block on pick-one
defer(coro, label=None)              # fire-and-forget, channel push on completion
@every(seconds)                      # periodic ticker, fn.cancel() to stop

Browser builtins (when repld[browser] is installed):

tab = await browser.get("*example.com*")  # find tab by URL glob
tab = await browser.open("https://...")   # open new tab
await browser.watch("*pattern*")          # auto-attach matching tabs

await tab.js("document.title")            # eval JS
await tab.fetch("/api/data")              # in-page fetch (inherits session)
await tab.click("#submit")                # trusted click
await tab.type_text("#search", "query")   # trusted typing
tab.network(url="*api*")                  # query captured traffic

Gists

Gists are Python modules in ./gists/ (project) or ~/.repld/gists/ (global) that wrap anything into a callable API — web apps via the browser, databases, graph stores, embedding indexes, internal services.

# gists/myapp.py
"""MyApp — accounts and transactions."""

class MyApp:
    def __init__(self, tab): self._tab = tab

    @classmethod
    async def connect(cls):
        from __main__ import browser
        tab = await browser.get("*myapp.com*")
        return cls(tab)

    async def accounts(self):
        return (await self._tab.fetch("/api/accounts"))["body"]

import myapp
app = await myapp.MyApp.connect()
await app.accounts()

Re-importing after edits auto-reloads. Gists can register MCP tools via __repld_tools__ — scaffold with repld gist <name>. Run repld help gists for details.

Browser

repld[browser] attaches to Chrome via CDP (--remote-debugging-port=9222). You log in normally; the agent sees your traffic, discovers the API surface, and works with your authenticated sessions.

tab = await browser.get("*salesforce*")
reqs = tab.network(url="*/api/*")           # discover API calls
auth = reqs[0].request_headers["Authorization"]  # extract auth

Body capture via Fetch interception means login flows, redirects, and CSRF exchanges are never lost. See docs/browser.md for the full design.

Scope

repld executes arbitrary Python in your project environment. It is a dev-time tool — never a runtime dependency. The IPC socket is localhost-only with user-only permissions.

Channels are a research-preview feature of Claude Code. The current integration uses --dangerously-load-development-channels server:repld.

License

MIT