GrantAi Memory

com.solonaiPublished by solonai.comv1.8.6Updated Oct 24, 2026

Persistent memory for AI agents. Infinite context with sub-millisecond recall.

The publisher's own server: Memory.fit

VerifiedSTDIODesktop onlyKnowledge & MemoryDeveloper Tools

Overview

AI-generated overview

GrantAi Memory is a local MCP server that gives AI assistants persistent, deterministic memory for storing and recalling exact content.

What you can do
Query memory for relevant context and store new content for later recall, optionally tagged with a speaker and filtered by agent. Import files, directories and git commit history into memory. Store session summaries, project state, code snippets and conversation turns for continuity. Check server health and status. Share one memory pool across multiple agents.
When to use it
Useful when you want an assistant to remember facts, project state or prior sessions across conversations, or when several agents should share a common knowledge base with exact, attributable recall instead of similarity search.
Requirements
Runs as a local process on the user's machine (desktop only) via a macOS/Linux/Windows package or a Docker image. The free edition needs no account or license key, but a one-time model download of about 650 MB is required; the full edition needs a GRANTAI_LICENSE_KEY. Docker installs need a persistent volume for memory data.
Before you install
The full edition requires a GRANTAI_LICENSE_KEY secret. The installer runs a remote script and registers the server with several AI tools, so review it before running. Stored memory persists locally and is shared by every agent pointed at the same instance or volume, so anything taught to it remains readable by those agents.

Installation

In SourceWeft

  1. Open GrantAi Memory in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

GrantAi

Deterministic Memory for AI
Local. Private. Secure.

Website • Download • Documentation

[GrantAi Demo]

Free Edition — one command, no account

The free edition is the full local memory engine as an MCP server: no email, no license key, no trial, and no network after the one-time model download (about 650 MB).

bash
# macOS / Linuxcurl -fsSL "https://solonai.com/grantai/install.sh?src=github" | bash
powershell
# Windows 10/11 (PowerShell)irm "https://solonai.com/grantai/install.ps1?src=github" | iex

The installer verifies both downloads against a published SHA-256 manifest, installs under ~/.grantai/free (Windows: %LOCALAPPDATA%\GrantAi\free), and registers the server with Claude Code, Claude Desktop, Cursor, Gemini CLI and Codex where they are installed. Any other MCP client takes the same command and arguments. Homebrew: brew install solonai/grantai/grantai-free. Air-gapped installs and the full list of what runs where are on the developers page.

The Problem

Every AI system today has the same flaw: it guesses instead of remembers.

RAG (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.

Ask 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.

This is the Retrieval Tax:

  • Re-retrieval — Same questions, same searches, same cost
  • Over-retrieval — 20 chunks when you need 3
  • Labor — Engineers tuning embeddings instead of building products
  • Risk — Approximate answers in domains that require precision

Enterprise AI spends 85% of compute on inference. Most of that is wasted on retrieving content that doesn't answer the question.

The Solution

GrantAi is deterministic memory for AI agents.

Instead 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.

RAGGrantAi
Returns similar contentReturns the exact content
10-20 chunks, hope one is right1-3 sentences, always right
Slows down as corpus growsMilliseconds regardless of size
No attributionFull audit trail
ApproximateDeterministic

Result: 97% reduction in tokens sent to the LLM. Faster responses. Lower cost. No hallucination from retrieval.

Why It Matters

  • Compliance — Exact citations, not paraphrased guesses
  • Multi-Agent — Shared memory across your AI workforce with speaker attribution
  • Cost — Pay for answers, not for searching
  • Security — 100% local, AES-256 encrypted, zero data egress

Quick Start

macOS / Linux (Native)

bash
# 1. Download from https://solonai.com/grantai/download# 2. Extract and install./install.sh
# 3. Restart your AI tool (Claude Code, Cursor, etc.)

Docker (All Platforms)

bash
docker pull ghcr.io/solonai-com/grantai-memory:1.8.6

Add to your Claude Desktop config (~/.config/Claude/claude_desktop_config.json):

json
{  "mcpServers": {    "grantai": {      "command": "docker",      "args": ["run", "-i", "--rm", "--pull", "always",               "-v", "grantai-data:/data",               "ghcr.io/solonai-com/grantai-memory:1.8.6"]    }  }}

Supported Platforms

PlatformMethodStatus
macOS (Apple Silicon)Native✅
Linux (x64)Native✅
WindowsNative✅
All PlatformsDocker✅

MCP Tools

GrantAi provides these tools to your AI:

ToolDescription
grantai_inferQuery memory for relevant context
grantai_teachStore content for future recall
grantai_learnImport files or directories
grantai_healthCheck server status
grantai_summarizeStore session summaries
grantai_projectTrack project state
grantai_snippetStore code patterns
grantai_gitImport git commit history
grantai_captureSave conversation turns for continuity

Multi-Agent Memory Sharing

Multiple agents can share knowledge through GrantAi's memory layer.

Basic shared memory (no setup required)

python
# Any agent storesgrantai_teach(    content="API rate limit is 100 requests/minute.",    source="api-notes")
# Any agent retrievesgrantai_infer(input="API rate limiting")

All agents read from and write to the same memory pool. No configuration needed.

With agent attribution (optional)

Use speaker to track which agent stored what, and from_agents to filter retrieval:

python
# Store with identitygrantai_teach(    content="API uses Bearer token auth.",    source="api-research",    speaker="researcher"  # optional)
# Retrieve from specific agentgrantai_infer(    input="API authentication",    from_agents=["researcher"]  # optional filter)

When to use speaker

ScenarioUse speaker?Why
Shared knowledge baseNoAll contributions equal, no filtering needed
Session continuityNoSame context, just persist and retrieve
Research → Code handoffYesCoder filters for researcher's findings only
Role-based trustYesSecurity agent's input treated differently

Framework integration

GrantAi works with any MCP-compatible client. Point your agents at the same GrantAi instance:

json
{  "mcpServers": {    "grantai": {      "command": "docker",      "args": ["run", "-i", "--rm", "--pull", "always",               "-v", "grantai-data:/data",               "ghcr.io/solonai-com/grantai-memory:1.8.6"]    }  }}

All agents using this config share the same memory volume (grantai-data).

Built By

GrantAi is built by Lawrence Grant, founder of SolonAI.

Background: Harvard, IBM, AI architecture and security work for Blackstone, Goldman Sachs, and Vanguard. Author of Mergers and Acquisitions Cybersecurity: The Framework For Maximizing Value.

Why We Built This

Read the full case for deterministic memory: Your AI Has Amnesia. You're Paying. Blame the Architecture.

Documentation

Support

License

Free to try. Pricing & Terms

Get Started →

Source: README.md at commit 3467360

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v1.8.6LatestSep 16, 2026

Score

40% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

    Declare a license in the repository or the registry entry.

Other Memory integrations

Memory.fit

fit.memoryOfficial publisher

Hosted Memory.fit server that lets an AI read your training, sleep, recovery and nutrition data and write back weekly plans, targets and meal plans.

Streamable HTTPWeb executableAuth
Location & LifestyleCredential requiredOct 29, 2026

pg.ddx.io: the PostgreSQL community's memory

io.ddx.pgOfficial publisher

pg.ddx.io: the PostgreSQL community's memory's hosted MCP server, published by pg.ddx.io; no sign-in required; Streamable HTTP; 29 tools.

Streamable HTTPWeb executable
Databases29 toolsOct 24, 2026

var.gg Naming Memory

gg.varOfficial publisher

Remote MCP server that suggests and searches code identifier names from domain concepts in Korean, Japanese, Chinese, Vietnamese and English, and exposes ERD…

Streamable HTTPWeb executable
Databases7 toolsOct 27, 2026

Counterparts

ai.counterparts

Local-first long-term memory for Claude Code and Claude Desktop: the assistant writes notes about you and your projects and recalls them in later sessions.

STDIODesktop only
Knowledge & MemoryNot validated yetNov 7, 2026

LAHZE - Persian AI Memory Platform

Published by GitHub account pedramgholizadeh

LAHZE - Persian AI Memory Platform's hosted MCP server; no sign-in required; Streamable HTTP; 13 tools.

Streamable HTTPWeb executable
Location & Lifestyle13 toolsNov 3, 2026