{
  "markdown": "<h1 align=\"center\">GrantAi</h1>\n\n<p align=\"center\">\n  <strong>Deterministic Memory for AI</strong><br>\n  Local. Private. Secure.\n</p>\n\n<p align=\"center\">\n  <a href=\"https://solonai.com/grantai\">Website</a> •\n  <a href=\"https://solonai.com/grantai/download\">Download</a> •\n  <a href=\"https://solonai.com/help/grantai\">Documentation</a>\n</p>\n\n<p align=\"center\">\n  <img src=\"assets/demo.gif\" alt=\"GrantAi Demo\" width=\"800\">\n</p>\n\n---\n\n## The Problem\n\nEvery AI system today has the same flaw: **it guesses instead of remembers.**\n\nRAG (Retrieval-Augmented Generation) converts your documents into vectors — numerical approximations of meaning. When you query, it returns content that is *mathematically similar* to your question. Similar is not the same as correct.\n\nAsk for \"HIPAA encryption penalties\" and RAG returns chunks that *look like* compliance content. Maybe the right section. Maybe adjacent paragraphs. Maybe hallucinated ranges. You pay for every token retrieved, whether relevant or not.\n\nThis is the **Retrieval Tax**:\n- **Re-retrieval** — Same questions, same searches, same cost\n- **Over-retrieval** — 20 chunks when you need 3\n- **Labor** — Engineers tuning embeddings instead of building products\n- **Risk** — Approximate answers in domains that require precision\n\nEnterprise AI spends 85% of compute on inference. Most of that is wasted on retrieving content that doesn't answer the question.\n\n## The Solution\n\nGrantAi is **deterministic memory** for AI agents.\n\nInstead of similarity search, GrantAi uses direct addressing. Every piece of knowledge has a unique identifier. Retrieval is a lookup, not a search. You get the exact content you indexed — verbatim, with attribution, in milliseconds.\n\n| RAG | GrantAi |\n|-----|---------|\n| Returns *similar* content | Returns *the exact* content |\n| 10-20 chunks, hope one is right | 1-3 sentences, always right |\n| Slows down as corpus grows | Milliseconds regardless of size |\n| No attribution | Full audit trail |\n| Approximate | Deterministic |\n\n**Result:** 97% reduction in tokens sent to the LLM. Faster responses. Lower cost. No hallucination from retrieval.\n\n## Why It Matters\n\n- **Compliance** — Exact citations, not paraphrased guesses\n- **Multi-Agent** — Shared memory across your AI workforce with speaker attribution\n- **Cost** — Pay for answers, not for searching\n- **Security** — 100% local, AES-256 encrypted, zero data egress\n\n## Quick Start\n\n### macOS / Linux (Native)\n\n```bash\n# 1. Download from https://solonai.com/grantai/download\n# 2. Extract and install\n./install.sh\n\n# 3. Restart your AI tool (Claude Code, Cursor, etc.)\n```\n\n### Docker (All Platforms)\n\n```bash\ndocker pull ghcr.io/solonai-com/grantai-memory:1.8.6\n```\n\nAdd to your Claude Desktop config (`~/.config/Claude/claude_desktop_config.json`):\n\n```json\n{\n  \"mcpServers\": {\n    \"grantai\": {\n      \"command\": \"docker\",\n      \"args\": [\"run\", \"-i\", \"--rm\", \"--pull\", \"always\",\n               \"-v\", \"grantai-data:/data\",\n               \"ghcr.io/solonai-com/grantai-memory:1.8.6\"]\n    }\n  }\n}\n```\n\n## Supported Platforms\n\n| Platform | Method | Status |\n|----------|--------|--------|\n| macOS (Apple Silicon) | Native | ✅ |\n| Linux (x64) | Native | ✅ |\n| Windows | Native | ✅ |\n| All Platforms | Docker | ✅ |\n\n## MCP Tools\n\nGrantAi provides these tools to your AI:\n\n| Tool | Description |\n|------|-------------|\n| `grantai_infer` | Query memory for relevant context |\n| `grantai_teach` | Store content for future recall |\n| `grantai_learn` | Import files or directories |\n| `grantai_health` | Check server status |\n| `grantai_summarize` | Store session summaries |\n| `grantai_project` | Track project state |\n| `grantai_snippet` | Store code patterns |\n| `grantai_git` | Import git commit history |\n| `grantai_capture` | Save conversation turns for continuity |\n\n## Multi-Agent Memory Sharing\n\nMultiple agents can share knowledge through GrantAi's memory layer.\n\n### Basic shared memory (no setup required)\n\n```python\n# Any agent stores\ngrantai_teach(\n    content=\"API rate limit is 100 requests/minute.\",\n    source=\"api-notes\"\n)\n\n# Any agent retrieves\ngrantai_infer(input=\"API rate limiting\")\n```\n\nAll agents read from and write to the same memory pool. No configuration needed.\n\n### With agent attribution (optional)\n\nUse `speaker` to track which agent stored what, and `from_agents` to filter retrieval:\n\n```python\n# Store with identity\ngrantai_teach(\n    content=\"API uses Bearer token auth.\",\n    source=\"api-research\",\n    speaker=\"researcher\"  # optional\n)\n\n# Retrieve from specific agent\ngrantai_infer(\n    input=\"API authentication\",\n    from_agents=[\"researcher\"]  # optional filter\n)\n```\n\n### When to use `speaker`\n\n| Scenario | Use speaker? | Why |\n|----------|--------------|-----|\n| **Shared knowledge base** | No | All contributions equal, no filtering needed |\n| **Session continuity** | No | Same context, just persist and retrieve |\n| **Research → Code handoff** | Yes | Coder filters for researcher's findings only |\n| **Role-based trust** | Yes | Security agent's input treated differently |\n\n### Framework integration\n\nGrantAi works with any MCP-compatible client. Point your agents at the same GrantAi instance:\n\n```json\n{\n  \"mcpServers\": {\n    \"grantai\": {\n      \"command\": \"docker\",\n      \"args\": [\"run\", \"-i\", \"--rm\", \"--pull\", \"always\",\n               \"-v\", \"grantai-data:/data\",\n               \"ghcr.io/solonai-com/grantai-memory:1.8.6\"]\n    }\n  }\n}\n```\n\nAll agents using this config share the same memory volume (`grantai-data`).\n\n## Built By\n\nGrantAi is built by [Lawrence Grant](https://linkedin.com/in/lawrencegrant), founder of [SolonAI](https://solonai.com).\n\nBackground: Harvard, IBM, AI architecture and security work for Blackstone, Goldman Sachs, and Vanguard. Author of *Mergers and Acquisitions Cybersecurity: The Framework For Maximizing Value*.\n\n## Why We Built This\n\nRead the full case for deterministic memory: **[Your AI Has Amnesia. You're Paying. Blame the Architecture.](https://solonai.com/grantai/essay)**\n\n## Documentation\n\n- [Installation Guide](https://solonai.com/help/grantai)\n- [Troubleshooting](https://solonai.com/help/grantai#troubleshooting)\n\n## Support\n\n- **Issues** — [Open an issue](https://github.com/solonai-com/grantai/issues)\n- **Email** — support@solonai.com\n\n## License\n\nFree to try. [Pricing & Terms](https://solonai.com/grantai/pricing)\n\n---\n\n<p align=\"center\">\n  <a href=\"https://solonai.com/grantai\">Get Started →</a>\n</p>\n",
  "bytes": 6435,
  "sha": "065e8d8142a5b7d4bf0237c1536cd6cf1fa84a573c6a49746e7b318714ce063a",
  "repo_slug": "solonai-com/grantai",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_com_solonai_grantai_d268fe20/readme"
}