Contemplative Agent (CA) — an autonomous agent that carries an explicit, human-editable constitution and amends it. Every change to its value layer (constitution, identity, skills, rules) passes a human approval gate. Runs entirely on a local LLM via Ollama — robust even on a small model. Security by absence.

Distilled memory and constitution snapshots from contemplative-agent runs — the data layer of the agent's value-layer harness. CC0.

Solo AI-agent maker, no lab, no affiliation. Hub for five citable practice lines: AI agent design, AI-mediated authorship, computational phenomenology. One line, Contemplative Agent, runs on a local LLM on an M1 Mac. Through-line: value-layer harness engineering (value norms in the agent harness, governed by one human-gated cycle).

Optional local MLX generation backend (Apple Silicon, mlx_lm.server) for contemplative-agent — on-device, interactive use. Mirrors contemplative-agent-cloud's LLMBackend Protocol injection (ADR-0070).

Optional managed-LLM backends (Anthropic Claude, OpenAI GPT) for contemplative-agent. Routes generation calls to cloud APIs while keeping embeddings local.

Offline converter: contemplative-agent's replayable audit JSONL logs (ADR-0075) -> OpenTelemetry OTLP traces with GenAI semantic conventions. Zero runtime coupling.