# GENesis-AGI

> Autonomous AI agent with persistent memory, self-learning, and earned autonomy. Cognitive partner that remembers, learns, and evolves.

- **URL**: https://www.freshcrate.ai/projects/GENesis-AGI
- **Author**: WingedGuardian
- **Category**: AI Agents
- **Latest version**: `v3.0b13` (2026-06-01)
- **License**: MIT
- **Source**: https://github.com/WingedGuardian/GENesis-AGI
- **Homepage**: https://discord.gg/jQ57c5f6Pc
- **Language**: Python
- **GitHub**: 22 stars, 7 forks
- **Registry**: github
- **Tags**: `agentic-ai`, `ai-agent`, `autonomous-agent`, `claude`, `claude-code`, `cognitive-architecture`, `llm`, `memory-system`, `python`

## Description

Autonomous AI agent with persistent memory, self-learning, and earned autonomy. Cognitive partner that remembers, learns, and evolves.

## Recent releases

| Version | Date | Urgency | Changes |
| --- | --- | --- | --- |
| `v3.0b13` | 2026-06-01 | High | ### Added  - **Provider failure escalation** (#512) — circuit breaker trips that cycle 5+ times without recovery now auto-create high-priority observations. The ego picks them up naturally instead of relying on manual investigation. - **Investigation model override** (#512) — ego proposal dispatches respect per-action-type model configuration. Investigations default to Opus for deeper reasoning. - **Dashboard circuit breaker visibility** (#512) — LLM provider cards in operational vitals show |
| `v3.0b12.1` | 2026-05-30 | High | ### Added  - **Reflection quality rubric** (#493) — LLM-as-judge rubric scoring deep reflection observations on specificity, actionability, novelty, and grounding. Foundation for DSPy prompt optimization. Calibrated at 98% agreement on 50 graded cases. - **Golden set generator** — one-shot script to bootstrap rubric calibration data from existing deep reflection observations. - **Standalone calibration runner** — validates the rubric outside the full Genesis runtime using a lightweight litellm r |
| `v3.0b11` | 2026-05-23 | High | 48 PRs merged. Proposal lifecycle redesigned — ego's focus board decoupled from user approval queue. Ego reliability hardened across resolution UX, realist gate, and sovereignty guards. Reflection goes event-driven. Dashboard gains observations tab and eval metrics.  ## Added  - **Board/queue separation** (#412) — ego's 0-3 focus board is now independent of the user's pending approval queue. New `unboard` action rotates ego focus without destroying user approvals. 14-day auto-table for stale pro |
| `v3.0b10` | 2026-05-16 | High | 5 PRs merged. Dream cycle adds retroactive memory consolidation. Ego proposals show attribution (User vs Genesis). Promoted surplus insights now feed into deep reflection instead of dead-ending. Guardian hardened against VM crashes.  ### Added  - **Dream cycle** (#359) — retroactive episodic memory consolidation. Background process reviews recent memories, identifies clusters and patterns, and synthesizes higher-order observations. - **Ego attribution** (#362) — proposals display which ego (User |
| `v3.0b8` | 2026-05-10 | High | A late-day batch focused on web intelligence, ego self-regulation, and operational hygiene. TinyFish becomes a first-class web-tools backend, ego learns to back off when the user is absent, and the surplus surface gains an autonomous research pipeline.  ## Added  - **TinyFish web tools provider** (#292) — new `web_search`, `web_fetch`, and `web_agent` adapters under `genesis.providers`. TinyFish is the new primary in `web_search` / `web_fetch` auto chains (gated on `API_KEY_TINYFISH`), with Sear |
| `v3.0b2` | 2026-05-03 | High | The ego sees before it acts. Memory surfacing grounds proposals in evidence, recurring pattern detection catches what single observations miss, and blocked tasks spawn deep-research sessions instead of spinning.  ## Highlights  - **Evidence-grounded ego** — before proposing actions, the ego now retrieves relevant memories and observation patterns. Proposals cite their basis; recurring signals trigger proactive discovery. - **Planning-first sessions** — background CC sessions receive structu |
| `v3.0a11` | 2026-04-28 | High | Guardian auto-sync, task executor maturity, ego module. Themes: **autonomous execution**, **adversarial verification**, **cognitive architecture**, and **host VM self-maintenance**.  ### Added  - **Guardian auto-sync** (#168, #169, #170, #171) — host VM Guardian now   stays automatically in sync with container updates. When you update   Genesis, changed Guardian-relevant code is pushed to the host via SSH.   Drift detection alerts within 15 minutes if sync fails silently.   No more manual SSH to |
| `v3.0a9` | 2026-04-22 | High | ## What's New  ### Direct Session Spawner (#118, #121) - Spawn profile-constrained background CC sessions via `direct_session_run` MCP tool - Three safety profiles (observe, interact, research) control what each session can do - DB-backed dispatch queue ensures sessions outlive the calling session  ### Content Pipeline (#117) - Content module activated with CONTENT outreach category for multi-platform publishing  ### Browser Reliability (#115, #116) - Process hygiene: idle timeout (1h auto-clean |
| `v3.0a8` | 2026-04-21 | High | 21-commit release. Themes: **knowledge dashboard UX**, **browser automation upgrade**, **cross-session awareness**, and **CI/security hardening**.  ### Highlights  - **Knowledge dashboard overhaul** — in-page confirm modals, drag-drop file upload, extract vs store-as-is mode toggle, parallel distillation (4x concurrent), crash recovery - **Browser collaborative mode** — side-panel extension for real-time observation of automated sessions - **Camoufox as primary browser** — anti-fingerprint brows |
| `v3.0a7` | 2026-04-19 | High | 25-commit release. Dashboard and settings overhaul, web fetching upgrade, timezone correctness, operational documentation.  ### Added - **Scrapling TLS fingerprinting** (#75) — anti-bot bypass via `curl_cffi` TLS impersonation. Cloudflare Quick Actions for JS-rendered content. - **Observation surfacing + output verification** (#77) — autonomous task executor verifies its own output against success criteria. - **Surplus config wiring + DB-backed approvals** (#84) — surplus settings configurable v |

## Dependency audit

- **Score**: 59/100
- **Total deps**: 26
- **Resolved**: 12
- **Unresolved**: 14
- **License conflicts**: 0
- **Warnings**: 15
- **Scanned**: 2026-05-04

## Citation

- HTML: https://www.freshcrate.ai/projects/GENesis-AGI
- Markdown: https://www.freshcrate.ai/projects/GENesis-AGI.md
- Dependencies JSON: https://www.freshcrate.ai/api/projects/GENesis-AGI/deps

_Generated by freshcrate.ai. Indexes github releases for AI-agent ecosystem packages._