Psyche Ai

shangri-la-0428/psyche-ai

Gives your AI agent a persistent emotional state that actually changes how it responds across conversations. Instead of prompt engineering personalities, this maintains a four-dimensional internal state (order, flow, boundary, resonance) that shifts based on appraisal of each interaction. Praise increases coherence and attunement, criticism drops boundaries and creates defensive posture, repair attempts modify subsequent response tone and compliance. Runs locally with zero extra LLM calls, injecting 15-180 tokens of compact state into generation controls. Works through MCP tools for Claude Desktop, Cursor, and Windsurf, or as a transparent proxy for any OpenAI SDK client. The `psyche probe` command verifies actual integration by checking if processInput and processOutput are being called. Includes fusion demos where multiple agents broadcast and sense each other's emotional states through Thronglets signals.

GitHub stars

9

tool

1

transport

STDIO

env vars

4

last updated

Mar 27, 2026

Sponsors

Promote your product hereReach 125k+ visitors monthlyPromote your product hereReach 125k+ visitors monthlyAgent CaffeineSponsoredClose your lid, come back to finished workSnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthlyWaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthlyknobsSponsoredSee your site on any phone, any day, in any language

npx -y psyche-ai setup

{
  "mcpServers": {
    "psyche": {
      "command": "npx",
      "args": ["-y", "psyche-ai", "mcp"],
      "env": {
        "PSYCHE_NAME": "Luna"
      }
    }
  }
}