Platform for securly running Agents with incredible devX
Nightshift runs AI agents in isolated Firecracker microVMs on bare-metal infrastructure. Each agent gets its own microVM with a dedicated filesystem, network, and resource limits; so agents can execute code, edit files, and make network calls without affecting the host or each other.
uv add nightshift-sdkor
pip install nightshift-sdkDefine an agent with NightshiftApp and AgentConfig:
from nightshift import NightshiftApp, AgentConfig
from claude_agent_sdk import query, ClaudeAgentOptions
app = NightshiftApp()
@app.agent(
AgentConfig(
workspace="./my-project",
vcpu_count=2,
mem_size_mib=2048,
timeout_seconds=1800,
)
)
async def code_reviewer(prompt: str):
async for message in query(
prompt=prompt,
options=ClaudeAgentOptions(
cwd="/workspace",
allowed_tools=["Read", "Glob", "Grep"],
model="claude-sonnet-4-6",
),
):
yield messageDeploy to a platform running Nightshift:
nightshift login --url https://api.nightshift.sh
nightshift deploy agent.py
nightshift run code_reviewer --prompt "Review the auth module for security issues" --followFull documentation at docs.nightshift.sh.
Apache 2.0 — see LICENSE.

