Open source. Self-deployed.
+Agents API
infrastructure.
On your terms.
+ Run Codex, Claude Code and MiniMax Code through one Core API and a shared Runtime. Deploy it on your infrastructure, with a Web console included.
+Your service. Your Runtime. Your model provider.
+import os
+from openai import OpenAI
+
+client = OpenAI(
+ base_url="http://localhost:8091/v1",
+ api_key=os.environ["PARSAR_CORE_API_KEY"],
+)
+
+for agent in client.beta.agents.list():
+ print(agent.id)
+ Use your Core API key and endpoint. This reads saved Agents; it does not call a model. Check coverage →
+01 / Architecture
One execution path.
Multiple native harnesses.
Core owns the API and execution state. The shared Runtime connects that contract to each native harness.
Parsar Core
Agents, Sessions & execution state
Runtime
Daemon, tools & workspace
Web console included. Connect to your Core through the same public API.
Native differences stay explicit. Supported capabilities depend on the harness and deployment profile.
02 / Installation
Start with a
local bundle.
Download or build a bundle for your host, extract it, then run the installer from that directory.
The default installs Core and the Web console, and prepares microsandbox and Runtime images. Core provisions the Runtime when you create an execution Session.
Installation guide03 / Before you deploy
Know what you are running.
The pinned Agents API is the compatibility target. Coverage is tracked by operation and workflow; full compatibility is not claimed.
Connect your first client
Configure your Core endpoint and authentication, then make your first API request.
Quickstart + 02Check protocol coverage
Review verified workflows, harness differences and known gaps against the pinned baseline.
Coverage & limitations + 03Operate your deployment
Self-deployment means owning service and container health, Runtime availability and recovery.
Operations guide +