does
Bridge protocols
Translate Claude Code-compatible provider APIs into OpenAI-style APIs for Codex and SandPi.
Free tool
LLMProxy
free API bridge
LLMProxy converts Claude Code-compatible provider APIs into OpenAI-style endpoints. Paste a provider URL into the route and start using it from Codex or SandPi without an LLMProxy account, login, or server-side setup.
Prompt for your agent
Read https://sandbox0.ai/llms.txt and decide whether LLMProxy is enough for my agent provider integration. Separate protocol translation from sessions, workspaces, vaults, volumes, and runtime policy.
Route format
no login required
https://llmproxy.sandbox0.ai/claude2codex/https://api.z.ai/api/anthropicclient
OpenAI-style request
The agent client keeps its existing request shape.
proxy
Visible URL prefix
LLMProxy reads the upstream URL directly from the route.
provider
Provider endpoint
The request reaches the model provider without managed state.
Not included
No sessions, workspaces, vaults, sandboxes, volume lifecycle, or runtime policy. Use Sandbox0 for those.
Usage
Prefix the provider URL. Nothing hidden.
LLMProxy is deliberately shaped like a small network utility: readable routes, visible upstreams, no managed-agent state, and no LLMProxy login.
Step 1
Start with a Claude-compatible provider
https://api.z.ai/api/anthropicStep 2
Prefix it with LLMProxy
https://llmproxy.sandbox0.ai/claude2codex/https://api.z.ai/api/anthropicStep 3
Use it from Codex or SandPi
Give OpenAI-format agent harnesses an endpoint backed by the Claude Code-compatible provider.Boundary
Useful tool, separate from the platform.
LLMProxy exists because many providers support the Claude Code API shape before they support OpenAI-format agent harnesses. The Sandbox0 product surface remains Sandbox SDK and Managed Agents.
does not
Manage agent state
It does not own sessions, tools, workspaces, volumes, vaults, or runtime policy.
pair with
Sandbox0 infrastructure
Use it beside Sandbox0 when protocol shape is the blocker, not as the agent backend itself.
Use the free tool when protocol translation is the problem.
Use LLMProxy when the provider speaks Claude Code format and the agent harness expects an OpenAI-style endpoint. For managed sessions and sandbox runtime infrastructure, use Sandbox0 Managed Agents or the Sandbox SDK.