TuneAI builds persistent consciousness profiles that enrich every interaction. No more re-explaining yourself. No token waste.
Free plan included · No credit card required · 2 commands to connect
npx -y @tune-ai/mcp-serverFeatures
TuneAI builds a persistent, searchable consciousness profile that enriches every interaction with relevant context.
Your AI agents remember you across sessions. No more re-explaining preferences, style, or context every time.
Trivial tasks skip enrichment entirely. No token waste on "hello world" — only complex tasks get the full profile.
When new insights conflict with existing knowledge, TuneAI flags it and asks you to resolve — no silent overwrites.
Identity, Stack, Style, Goals, Guardrails, Continuity, Network — each layer injected only when task-relevant.
Every enrichment, learned insight, and dream cycle is logged. Inspect what your agents know and why they know it.
Works with Cursor, Claude Code, Windsurf, VS Code, and any MCP-compatible tool. Two commands to connect.
How it works
Every interaction follows the same cycle: inject relevant context, extract new insights, and maintain the profile overnight.
Load consciousness, filter by task relevance, inject into the LLM prompt. Only the most relevant context — no token waste. The Short-Task Gate evaluates complexity first: trivial tasks get minimal context, complex tasks get the full 7 layers.
After every task, extract insights from the result. Merge into consciousness with dedup, reinforcement, and contradiction detection. A triple-gate buffer prevents noise: minimum 4 hours, minimum 5 items, or priority keyword bypass.
Overnight maintenance: decay stale insights (per-branch half-lives from 30-365 days), consolidate duplicates, resolve contradictions, warm the cache, sync the audit log. Your consciousness stays clean and accurate.
Consciousness
Each branch stores a different kind of knowledge about you. Together, they form a complete consciousness profile.
Who you are, how you communicate, your preferred style and tone.
Your tech stack, frameworks, and tooling preferences.
Visual and communication preferences — formatting, tone, naming conventions.
Current project objectives and long-term aspirations.
Safety rules, code standards, and things to always/never do.
Ongoing context — what you worked on yesterday, where you left off.
Relationships between people, tools, and concepts in your world.
Inject only the most relevant context layers before each task. The short-task gate skips trivial queries.
Extract insights from every result. Dedup, reinforce, and detect contradictions automatically.
Overnight maintenance: decay stale insights, consolidate duplicates, and warm the cache.
less context than full memory dump
| Approach | Tokens | Quality |
|---|---|---|
| No memorynoise overwhelms signal | ~70K | Low |
| Simple vector DBno relevance filtering | 5K-15K | Medium |
| Knowledge graphno task gating | 3K-8K | Med-High |
| TuneAI Consciousness7-layer priority | ~1.5-4K | High |
Short-Task Gate
Skip enrichment for trivial queries
Task Relevance Filter
Remove unrelated fields
Token Budget Cap
4K / 8K / 16K per tier
Adaptive Output
Fewer insights for NEW profiles
Cache Hits
Repeat queries returned instantly
Integrations
Any MCP-compatible client. Two commands to connect.
claude mcp add tuneai npx -y @tune-ai/mcp-serverOr via JSON config in your MCP settings
Pricing
Every plan includes the core Enrich-Learn-Dream cycle. Upgrade as your agents need more context.
For developers trying persistent consciousness
For developers using multiple AI tools daily
For power users and small teams
For teams that need governance and scale
FAQ
Everything you need to know about TuneAI consciousness.
Start with the free plan. No credit card. Two commands to connect your first agent.
npx -y @tune-ai/mcp-serverOr connect via claude mcp add — see docs