Skip to main content
Condor uses ACP (Agent Client Protocol) to connect Trading Agents to LLMs. ACP is a standardized protocol for agent-editor communication that enables interoperability between clients and AI agents.

​
Supported LLMs

ProviderModels
AnthropicClaude Opus, Sonnet, Haiku
GoogleGemini Pro, Gemini Flash
OpenAICodex, GPT-4, GPT-4o

​
How ACP Works

ACP provides a standardized protocol for agent-editor communication, similar to how the Language Server Protocol (LSP) standardized language server integration.
  • Local deployment: Agents operate as sub-processes, exchanging data via JSON-RPC over standard input/output
  • Remote deployment: Agents can be cloud-hosted, communicating through HTTP or WebSocket connections
  • Interoperability: Any LLM implementing ACP can power the reasoning layer for Trading Agents

​
Agent Portability

Agents are defined as structured Markdown files, enabling:
  • Version control: Track changes with git
  • Sharing: Share agent configurations with your team
  • Auditing: Every session logs each turn as a structured snapshot
  • Cross-platform: Start a conversation on Telegram, continue in Claude Code

​
Using /agent Command

The /agent command in Telegram lets you connect to LLMs and build autonomous Trading Agents.
See ACP documentation for the full protocol specification.