Exocortex

MCP文档知识

提供方 fuwasegu

文档知识 HTTP · 远程 会话登录 社区 / 第三方 CursorCodexZed
查看接入方式

xocortex — an MCP server that gives your AI persistent memory across ALL your projects.

一句话选型结论

按资料字段整理

Exocortex 属于文档知识类 MCP 服务器,HTTP · 远程,会话登录,可直接接入 Cursor、Codex、Zed,有可直接复制的远程端点配置。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类文档知识 传输方式HTTP · 远程 鉴权会话登录 部署方式远程部署
GitHub 星标 未标注 仓库热度,无公开仓库时显示未标注
工具数 未标注 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

文档知识
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量未标注暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

支持客户端

适用场景

mcpai memoryknowledge graph

安装与配置

HTTP · 远程 会话登录
连远程 HTTPS 端点,通常为 Streamable HTTP
uvx --from git+https://github.com/fuwasegu/exocortex exocortex

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    确认你的客户端支持远程端点,并拿到服务地址。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后随便调一个只读工具,确认能返回结果,再接入正式流程。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
文档检索问答

在你的文档库里按问题找答案。

知识库读写

把新资料写进知识库,随用随取。

内容整理

把零散资料归纳成结构化笔记。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 已适配 Cursor、Codex、Zed,换客户端不用重新研究配置。

先想清楚的地方

  • 远程端点依赖网络与服务方可用性,注意密钥与调用额度。
  • 使用前要准备好会话登录对应的凭据,并按最小权限配置。
  • 资料没标注工具数量,实际能力要装好后确认。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在文档知识场景里让模型直接调用外部能力的人。
  • 已经在用 Cursor、Codex 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 不方便管理密钥或第三方授权的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
Exocortex当前 未标注 HTTP · 远程 远程部署 有 未标注
dify 未标注 stdio · 本地进程 本地部署 有 2026-04-11
open-webui 未标注 stdio · 本地进程 本地部署 有 2026-04-11
JeecgBoot 未标注 stdio · 本地进程 本地部署 有 2026-04-11
Cherry Studio 未标注 未标注 未标注 无 2026-05-17
mempalace 未标注 stdio · 本地进程 本地部署 有 2026-04-11

服务器档案

文档知识
字段来自项目仓库与公开索引,中文说明由本站标注
提供方fuwasegu上游标注的维护方 fuwasegu
分类文档知识本站按上游分类展示
连接方式HTTP · 远程连远程 HTTPS 端点,通常为 Streamable HTTP
鉴权方式会话登录远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时未标注 · Cross-platform本地安装时用到的包标识与运行时
部署类型远程部署上游给出的部署形态
最近核验未标注上游最后核验时间

详细介绍

30.3 KB
来自项目仓库与上游目录的原文,未做删改

Exocortex 🧠

> "Extend your mind." - Your External Brain

日本語版はこちら (Japanese)


Exocortex is a local MCP (Model Context Protocol) server that acts as a developer's "second brain."

It persists development insights, technical decisions, and troubleshooting records, allowing AI assistants (like Cursor) to retrieve contextually relevant memories when needed.

Why Exocortex?

🌐 Cross-Project Knowledge Sharing

Unlike tools that store data per-repository (e.g., .serena/ in each project), Exocortex uses a single, centralized knowledge store.```
Traditional approach (per-repository):
project-A/.serena/ ← isolated knowledge
project-B/.serena/ ← isolated knowledge
project-C/.serena/ ← isolated knowledge

Exocortex approach (centralized):
~/.exocortex/data/ ← shared knowledge across ALL projects
├── Insights from project-A
├── Insights from project-B
└── Insights from project-C
↓
Cross-project learning!


**Benefits:**
- 🔄 **Knowledge Transfer**: Lessons learned in one project are immediately available in others
- 🏷️ **Tag-based Discovery**: Find related memories across projects via shared tags
- 📈 **Cumulative Learning**: Your external brain grows smarter over time, not per project
- 🔍 **Pattern Recognition**: Discover common problems and solutions across your entire development history

## Features

- 🔒 **Fully Local**: All data and AI processing stays on your machine. Privacy guaranteed.
- 🔍 **Semantic Search**: Find memories by meaning, not just keywords.
- 🕸️ **Knowledge Graph**: Maintains relationships between projects, tags, and memories with explicit links.
- 🔗 **Memory Links**: Connect related memories to build a traversable knowledge network.
- ⚡ **Lightweight & Fast**: Uses embedded KùzuDB and lightweight fastembed models.
- 🧠 **Memory Dynamics**: Smart recall based on recency and frequency—frequently accessed memories surface higher.
- 🔥 **Frustration Indexing**: Prioritize "painful memories"—debugging nightmares get boosted in search results.
- 🖥️ **Web Dashboard**: Beautiful cyberpunk-style UI for browsing memories, monitoring health, and visualizing the knowledge graph.

## 📚 Usage Guide

**→ See the full usage guide**

- Tool reference with use cases
- Practical workflows
- Prompting tips
- Tips & Tricks

## Installation```bash
# Clone the repository
git clone https://github.com/fuwasegu/exocortex.git
cd exocortex

# Install dependencies with uv
uv sync

Usage

Starting the Server```bash

uv run exocortex


### Cursor Configuration

Add the following to your `~/.cursor/mcp.json`:

#### Option 1: Direct from GitHub (Recommended)

Auto-updates when uvx cache expires. No manual `git pull` needed.```json
{
  "mcpServers": {
    "exocortex": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/fuwasegu/exocortex", "exocortex"]
    }
  }
}

Option 2: Local Installation

For development or customization.```json
{
"mcpServers": {
"exocortex": {
"command": "uv",
"args": ["--directory", "/path/to/exocortex", "run", "exocortex"]
}
}
}


> **Note:** Your data is stored in `~/.exocortex/` and is preserved regardless of which option you choose.

#### Option 3: Proxy Mode (Multiple Cursor Instances - Recommended)

**Use this method if you want to use Exocortex from multiple Cursor windows simultaneously.**

KùzuDB doesn't support concurrent writes from multiple processes. With the stdio approach where each Cursor instance spawns its own server process, lock conflicts occur. Proxy mode automatically starts a single SSE server in the background, and each Cursor instance connects via proxy.```json
{
  "mcpServers": {
    "exocortex": {
      "command": "uvx",
      "args": [
        "--from", "git+https://github.com/fuwasegu/exocortex",
        "exocortex",
        "--mode", "proxy",
        "--ensure-server"
      ]
    }
  }
}

How it works:
1. First Cursor starts Exocortex → SSE server automatically starts in background
2. Subsequent Cursors connect to the existing SSE server
3. All Cursors share the same server → No lock conflicts!

> Note: No manual server startup required. The --ensure-server option automatically starts the server if it's not running.

Option 4: Manual Server Management (Advanced)

If you prefer to manage the server manually:

Step 1: Start the server```bash

Start the server in a terminal (can also run in background)

uv run --directory /path/to/exocortex exocortex --transport sse --port 8765


**Step 2: Configure Cursor**```json
{
  "mcpServers": {
    "exocortex": {
      "url": "http://127.0.0.1:8765/mcp/sse"
    }
  }
}

> Bonus: With this setup, you can also access the web dashboard at http://127.0.0.1:8765/

> Tip: To auto-start the server on system boot, use launchd on macOS or systemd on Linux.

MCP Tools

Basic Tools

Tool Description
exo_ping Health check to verify server is running
exo_store_memory Store a new memory
exo_recall_memories Recall relevant memories via semantic search
exo_list_memories List stored memories with pagination
exo_get_memory Get a specific memory by ID
exo_delete_memory Delete a memory
exo_get_stats Get statistics about stored memories

Advanced Tools

Tool Description
exo_link_memories Create a link between two memories
exo_unlink_memories Remove a link between memories
exo_update_memory Update content, tags, or type of a memory
exo_explore_related Discover related memories via graph traversal
exo_get_memory_links Get all outgoing links from a memory
exo_trace_lineage 🕰️ Trace the evolution/lineage of a memory (temporal reasoning)
exo_curiosity_scan 🤔 Scan for contradictions, outdated info, and knowledge gaps
exo_analyze_knowledge Analyze knowledge base health and get improvement suggestions
exo_sleep Trigger background consolidation (deduplication, orphan rescue, auto-linking)
exo_consolidate Extract abstract patterns from memory clusters
exo_compact 🧹 Rebuild the database to reclaim disk space (see Compaction)

🤖 Knowledge Autonomy

Exocortex automatically improves your knowledge graph! When you store a memory, the system:

  1. Suggests Links: Finds similar existing memories and suggests connections
  2. Detects Duplicates: Warns if the new memory is too similar to an existing one
  3. Identifies Patterns: Recognizes when a success might resolve a past failure```json
    // Example exo_store_memory response with suggestions
    {
    "success": true,
    "memory_id": "...",
    "suggested_links": [
    {
    "target_id": "existing-memory-id",
    "similarity": 0.78,
    "suggested_relation": "extends",
    "reason": "High semantic similarity; may be an application of this insight"
    }
    ],
    "insights": [
    {
    "type": "potential_duplicate",
    "message": "This memory is very similar (94%) to an existing one.",
    "suggested_action": "Use exo_update_memory instead"
    }
    ]
    }

### 🧠 Automatic Memory Consolidation

**Like human sleep consolidates memories, Exocortex prompts the AI to organize after storing.**

When `exo_store_memory` succeeds, the response includes `next_actions` that guide the AI to:

1. **Link high-similarity memories** (similarity ≥ 0.7)
2. **Handle duplicates and contradictions**
3. **Run periodic health checks** (every 10 memories)```json
// Example response with next_actions
{
  "success": true,
  "memory_id": "abc123",
  "summary": "...",
  "consolidation_required": true,
  "consolidation_message": "🧠 Memory stored. 2 consolidation action(s) required.",
  "next_actions": [
    {
      "action": "link_memories",
      "priority": "high",
      "description": "Link to 2 related memories",
      "details": [
        {
          "call": "exo_link_memories",
          "args": {
            "source_id": "abc123",
            "target_id": "def456",
            "relation_type": "extends",
            "reason": "High semantic similarity"
          }
        }
      ]
    },
    {
      "action": "analyze_health",
      "priority": "low",
      "description": "Run knowledge base health check",
      "details": { "call": "exo_analyze_knowledge" }
    }
  ]
}

Expected Flow:```
User: "Remember this insight"
↓
AI: exo_store_memory() → receives next_actions
↓
AI: exo_link_memories() for each high-priority action
↓
AI: "Stored and linked to 2 related memories."


> ⚠️ **Important Limitation**: Execution of `next_actions` is at the AI agent's discretion. While the server strongly instructs consolidation via `SERVER_INSTRUCTIONS` and `consolidation_required: true`, **execution is NOT 100% guaranteed**. This is an inherent limitation of the MCP protocol—servers can only suggest, not force actions. In practice, most modern AI assistants follow these instructions, but they may be skipped during complex conversations or when competing with other tasks.

### Relation Types for `exo_link_memories`

| Type | Description |
|------|-------------|
| `related` | Generally related memories |
| `supersedes` | This memory updates/replaces the target |
| `contradicts` | This memory contradicts the target |
| `extends` | This memory extends/elaborates the target |
| `depends_on` | This memory depends on the target |
| `evolved_from` | This memory evolved from the target (temporal reasoning) |
| `rejected_because` | This memory was rejected due to the target |
| `caused_by` | This memory was caused by the target |

### Temporal Reasoning with `exo_trace_lineage`

Trace the **lineage of decisions and knowledge** over time. Understand WHY something became the way it is.

| Parameter | Description | Example |
|-----------|-------------|---------|
| `memory_id` | Starting memory ID | `"abc123"` |
| `direction` | `"backward"` (find ancestors) or `"forward"` (find descendants) | `"backward"` |
| `relation_types` | Relations to follow | `["evolved_from", "caused_by"]` |
| `max_depth` | Maximum traversal depth | `10` (default) |

**Example: Understanding Why a Decision Was Made**```
Current Architecture Decision
    │
    ▼ trace_lineage(direction="backward")
    │
    ├─ [depth 1] Previous Design (evolved_from)
    │      "Switched from monolith to microservices"
    │
    └─ [depth 2] Original Problem (caused_by)
           "Scaling issues with single database"

Usage:```
AI: exo_trace_lineage(memory_id="current-decision", direction="backward")
↓
Result: Shows the evolution chain of how the current decision came to be


**Use Cases:**
- 🔍 **Architecture archaeology**: "Why did we choose this approach?"
- 🐛 **Root cause analysis**: "What led to this bug?"
- 📚 **Knowledge evolution**: "How has our understanding changed?"

### Curiosity Engine with `exo_curiosity_scan`

The Curiosity Engine actively **questions your knowledge base** like a curious human would. It scans for inconsistencies, finds unlinked memories, and generates questions to improve knowledge quality.

**What it detects:**

| Category | Description | Example |
|----------|-------------|---------|
| 🔴 **Contradictions** | Memories that conflict with each other | Success vs Failure on same topic |
| 📅 **Outdated Info** | Old knowledge that may need review | Memories superseded but not linked |
| 🔗 **Suggested Links** | Unlinked memories that should be connected | Memories sharing tags, context, or high similarity |
| ❓ **Questions** | Human-like questions about your knowledge | "Is this still valid?" |

**Suggested Links Detection Strategies:**

| Strategy | Confidence | Description |
|----------|------------|-------------|
| **Tag Sharing** | High (0.7+) | Memories sharing 2+ tags are likely related |
| **Context Sharing** | Medium (0.6) | Same project + same type (insight/decision) |
| **Semantic Similarity** | High (0.7+) | High vector similarity (>70%) but not linked |

**Example Output:**```json
{
  "contradictions": [
    {
      "memory_a_summary": "Caching approach works perfectly",
      "memory_b_summary": "Caching approach failed badly",
      "reason": "success vs failure on same topic",
      "confidence": 0.85
    }
  ],
  "suggested_links": [
    {
      "source_summary": "Database optimization technique",
      "target_summary": "Query performance improvement",
      "reason": "Share 3 tags: database, performance, optimization",
      "link_type": "tag_shared",
      "confidence": 0.8,
      "suggested_relation": "related"
    }
  ],
  "outdated_knowledge": [],
  "questions": [
    "🤔 These memories seem to contradict. Are both still valid?",
    "🔗 Found unlinked related memories. Link them to strengthen the graph?"
  ],
  "next_actions": [
    {
      "action": "create_link",
      "priority": "medium",
      "details": {
        "call": "exo_link_memories",
        "args": {
          "source_id": "...",
          "target_id": "...",
          "relation_type": "related"
        }
      }
    }
  ]
}

Usage:```
AI: exo_curiosity_scan(context_filter="my-project")
↓
Result: Report of issues, suggested links, and questions
↓
AI: Executes next_actions to create links
↓
Result: Knowledge graph becomes richer and more connected!


**Use Cases:**
- 🔍 **Knowledge audit**: "Are there any contradictions in my knowledge?"
- 🔗 **Graph enrichment**: "Find unlinked memories that should be connected"
- 🧹 **Quality maintenance**: "What needs to be cleaned up?"
- 💡 **Discovery**: "What questions should I be asking about my knowledge?"

### Optional: Enhanced Sentiment Analysis

For higher accuracy contradiction detection in `exo_curiosity_scan`, you can enable BERT-based sentiment analysis:```bash
# Local installation
pip install exocortex[sentiment]
# or
uv sync --extra sentiment
``````json
// mcp.json with sentiment support
{
  "mcpServers": {
    "exocortex": {
      "command": "uvx",
      "args": [
        "--from", "exocortex[sentiment] @ git+https://github.com/fuwasegu/exocortex",
        "exocortex", "--mode", "proxy", "--ensure-server"
      ]
    }
  }
}

> Note: Adds ~2.5GB of dependencies (PyTorch + Transformers). The default keyword-based detection works well for most cases and supports both English and Japanese.

Environment Variables

Variable Default Description
EXOCORTEX_DATA_DIR ~/.exocortex Database storage directory
EXOCORTEX_LOG_LEVEL INFO Logging level (DEBUG/INFO/WARNING/ERROR)
EXOCORTEX_EMBEDDING_MODEL sentence-transformers/all-MiniLM-L6-v2 Embedding model to use
EXOCORTEX_TRANSPORT stdio Transport mode (stdio/sse/streamable-http)
EXOCORTEX_HOST 127.0.0.1 Server bind address (for HTTP modes)
EXOCORTEX_PORT 8765 Server port number (for HTTP modes)

Architecture

Stdio Mode (Default)```

┌─────────────────┐ stdio ┌─────────────────────────────┐
│ AI Assistant │ ◄──────────► │ Exocortex MCP │
│ (Cursor) │ MCP │ │
└─────────────────┘ │ ┌─────────┐ ┌──────────┐ │
│ │ Tools │ │ Embedding│ │
│ │ Handler │ │ Engine │ │
│ └────┬────┘ └────┬─────┘ │
│ │ │ │
│ ┌────▼────────────▼─────┐ │
│ │ KùzuDB │ │
│ │ (Graph + Vector) │ │
│ └────────────────────────┘ │
└─────────────────────────────┘


### HTTP/SSE Mode (Multiple Instances)```
┌─────────────────┐
│  Cursor #1      │──────┐
└─────────────────┘      │
                         │  HTTP   ┌─────────────────────────────┐
┌─────────────────┐      ├────────►│       Exocortex MCP         │
│  Cursor #2      │──────┤   SSE   │      (Standalone)           │
└─────────────────┘      │         │                             │
                         │         │  ┌─────────┐  ┌──────────┐  │
┌─────────────────┐      │         │  │ Tools   │  │ Embedding│  │
│  Cursor #3      │──────┘         │  │ Handler │  │  Engine  │  │
└─────────────────┘                │  └────┬────┘  └────┬─────┘  │
                                   │       │            │        │
                                   │  ┌────▼────────────▼─────┐  │
                                   │  │       KùzuDB          │  │
                                   │  │  (Graph + Vector)     │  │
                                   │  └────────────────────────┘  │
                                   └─────────────────────────────┘

Knowledge Graph Structure```

Memory ─── ORIGINATED_IN ──► Context (project)
Memory ─── TAGGED_WITH ────► Tag
Memory ─── RELATED_TO ─────► Memory (with relation type)


### Memory Dynamics

Exocortex implements a **Memory Dynamics** system inspired by human cognition. Memories have "lifespan" and "strength" that affect search results:

**Hybrid Scoring Formula:**```
Score = (S_vec × w_vec) + (S_recency × w_recency) + (S_freq × w_freq) + (S_frustration × w_frustration)
Component Description Default Weight
S_vec Vector similarity (semantic relevance) 0.50
S_recency Recency score (exponential decay: e^(-λ×Δt)) 0.20
S_freq Frequency score (log scale: log(1 + count)) 0.15
S_frustration Frustration score (painful memory boost) 0.15

How it works:
- Every time a memory is recalled, its last_accessed_at and access_count are updated
- Frequently accessed memories gain higher S_freq scores
- Recently accessed memories gain higher S_recency scores
- Painful memories (debugging nightmares) get higher S_frustration scores for priority
- Old, unused memories naturally decay but remain searchable

This creates an intelligent recall system where:
- 📈 Important memories (frequently used) stay prominent
- ⏰ Recent context is prioritized
- 🔥 Painful memories are never forgotten—to avoid repeating mistakes
- 🗃️ Old memories gracefully fade but don't disappear

Frustration Indexing (Somatic Marker Hypothesis)

Based on the neuroscience insight that "painful memories are prioritized in decision-making", Exocortex automatically boosts the importance of debugging struggles and hard-won solutions.

Usage:```python

Explicitly mark as a painful memory

exo_store_memory(
content="Spent 3 hours debugging KùzuDB lock issues. Root cause was...",
context_name="exocortex",
tags=["bug", "kuzu"],
is_painful=True, # ← Important!
time_cost_hours=3.0, # ← Record time spent
)


**Auto-detection:**
Even without `is_painful`, frustration level is auto-detected from content:

- 😓 **Low** (0.2-0.4): "tricky", "weird", "workaround"
- 🔥 **Medium** (0.4-0.6): "finally", "bug", "hours"
- 🔥🔥 **High** (0.6-0.8): "stuck", "frustrated"
- 🔥🔥🔥 **Extreme** (0.8-1.0): "nightmare", "impossible", "hell"

**Search results:**```json
{
  "memories": [
    {
      "id": "...",
      "summary": "KùzuDB lock issue resolution",
      "frustration_score": 0.85,
      "pain_indicator": "🔥🔥🔥",   // ← Visual emphasis
      "time_cost_hours": 3.0
    }
  ]
}

Sleep/Dream Mechanism

Like human sleep consolidates memories, Exocortex has a background consolidation process that organizes your knowledge graph:```
┌─────────────────────────────────────────────────────────────┐
│ exo_sleep() called │
└─────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ Dream Worker (Detached Process) │
│ ┌──────────────────────────────────────────────────────┐ │
│ │ 1. Deduplication │ │
│ │ - Find memories with similarity >= 95% │ │
│ │ - Link newer → older with 'related' relation │ │
│ ├──────────────────────────────────────────────────────┤ │
│ │ 2. Orphan Rescue │ │
│ │ - Find memories with no tags and no links │ │
│ │ - Link to most similar memory with 'related' │ │
│ ├──────────────────────────────────────────────────────┤ │
│ │ 3. Auto-linking (High Confidence Only) │ │
│ │ - Tag sharing: 3+ shared tags → 'related' │ │
│ │ - Semantic: 80%+ similarity → 'related' │ │
│ ├──────────────────────────────────────────────────────┤ │
│ │ 4. Pattern Mining (Phase 2) │ │
│ │ - Extract common patterns from memory clusters │ │
│ └──────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────┘


**Usage:**```
AI: "I've completed the task. Let me consolidate the knowledge base."
    ↓
AI: exo_sleep() → Worker spawns in background
    ↓
AI: "Consolidation process started. Your knowledge graph will be optimized."

Key Features:
- 🔄 Non-blocking: Returns immediately, consolidation runs in background
- 🔐 Safe: Uses file locking to avoid conflicts with active sessions
- 📊 Logs: Enable logging with enable_logging=True to track progress

> ℹ️ Proxy Mode: Since v1.1.0 exo_sleep works in proxy mode. The server now hands its database lock over before spawning the worker, and releases the lock automatically after a few seconds of inactivity.
>
> This is best-effort rather than guaranteed: if the server is busy serving requests it may reacquire the database before the worker (which loads an embedding model first) gets in. If the worker reports a lock timeout, simply retry when things are quiet.

Compaction

If your database directory is far larger than the data it holds — hundreds of megabytes for a few thousand memories — it was bloated by a write-path bug present before v1.1.0. Every write query ran a full storage checkpoint, and each checkpoint rewrote relationship column chunks without reclaiming pages (~386 KiB of file growth per relationship). Badly affected databases eventually fail to read the bloated column at all:```
RuntimeError: Buffer manager exception: Unable to allocate memory!


v1.1.0 fixes the cause, but an already-bloated database does not shrink on its own. Rebuild it:```bash
# As a CLI (the server must not be running)
uvx --from git+https://github.com/fuwasegu/exocortex exocortex --mode compact

# Keep no backup (the pre-compaction copy is kept under ~/.exocortex/backups/ by default)
uvx --from git+https://github.com/fuwasegu/exocortex exocortex --mode compact --no-backup

Or ask your AI assistant to call exo_compact.

The original database is moved to ~/.exocortex/backups/ and only replaced once the rebuild fully succeeds. Relationship text that is already unreadable cannot be recovered — those columns are reported and rebuilt empty so the rest of the graph survives.

> EXPORT DATABASE is not a substitute: it runs out of memory on exactly the databases that need compacting.

Pattern Abstraction (Concept Formation)

Exocortex can extract abstract patterns from concrete memories, creating a hierarchical knowledge structure:```
┌─────────────────────────────────────────────────────────────────┐
│ Pattern Layer │
│ ┌───────────────────────────────────────────────────────────┐ │
│ │ "Always use connection pooling for database connections" │ │
│ │ Confidence: 0.85 | Instances: 5 │ │
│ └───────────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────┘
▲ INSTANCE_OF ▲ INSTANCE_OF
┌────────────┴────────────────┴────────────┐
┌──────┴──────┐ ┌───────────┐ ┌───────────┐ ┌──┴────────┐
│ Memory #1 │ │ Memory #2 │ │ Memory #3 │ │ Memory #4 │
│ PostgreSQL │ │ MySQL │ │ Redis │ │ MongoDB │
│ pooling fix │ │ pool size │ │ conn reuse│ │ pool leak │
└─────────────┘ └───────────┘ └───────────┘ └───────────┘
Memory Layer (Concrete)


**Usage:**```
AI: exo_consolidate(tag_filter="database") → Extracts patterns from database-related memories
    ↓
Result: "Created 2 patterns from 8 memories"

Benefits:
- 🎯 Generalization: Discover rules that apply across specific cases
- 🔍 Meta-learning: Find what works (and what doesn't) across projects
- 📈 Confidence Building: Patterns get stronger as more instances are linked

Web Dashboard

Exocortex includes a beautiful web dashboard for visualizing and managing your knowledge base.

Accessing the Dashboard

🚀 If You're Using Proxy Mode (Recommended)

No terminal commands needed! When using proxy mode (--mode proxy --ensure-server) with Cursor, the SSE server is automatically running in the background.

Just open this in your browser:```
http://127.0.0.1:8765/
``````
Cursor starts
↓
Proxy mode → SSE server auto-starts (port 8765)
↓
├─ MCP: http://127.0.0.1:8765/mcp/sse ← Used by Cursor
└─ Dashboard: http://127.0.0.1:8765/ ← Just open in browser!


#### Starting the Server Manually

If you want to view the dashboard without using Cursor:```bash
# Start SSE server (includes dashboard)
uv run exocortex --transport sse --port 8765

URLs:
- Dashboard: http://127.0.0.1:8765/
- MCP SSE: http://127.0.0.1:8765/mcp/sse

Dashboard Features

Tab Description
Overview Statistics, contexts, tags, and knowledge base health score
Memories Browse, filter, and search memories with pagination
Dream Log Real-time streaming log of background consolidation processes
Graph Visual knowledge graph showing memory connections

Screenshots

Overview Tab
- Total memories count by type (Insights, Successes, Failures, Decisions, Notes)
- Context and tag clouds for quick navigation
- Health score with improvement suggestions

Memories Tab
- Filter by type (Insight/Success/Failure/Decision/Note)
- Filter by context (project)
- Click any memory to see full details and links

Graph Tab
- Interactive node visualization
- Color-coded by memory type:
- 🔵 Cyan: Insights
- 🟠 Orange: Decisions
- 🟢 Green: Successes
- 🔴 Red: Failures
- Lines show RELATED_TO connections between memories

Standalone Dashboard Mode

You can also run the dashboard separately on a different port:```bash
uv run exocortex --mode dashboard --dashboard-port 8766


> **Note:** In standalone mode, the dashboard connects to the same database but doesn't include the MCP server.

## Documentation

- Design Document - System design and specifications
- Graph Architecture - How the knowledge graph works

## Development```bash
# Install dependencies
uv sync

# Run tests
uv run pytest

# Run with debug logging
EXOCORTEX_LOG_LEVEL=DEBUG uv run exocortex

License

MIT License

常见问题

8 问
答案由本站依据条目字段整理,不引用上游原文
Exocortex 怎么安装?

在客户端按下面的安装命令接入:uvx --from git+https://github.com/fuwasegu/exocortex exocortex。装完在客户端配置里引用同一个包或端点。

Exocortex 怎么在 Cursor、Codex、Zed 里配置?

它的传输方式是HTTP · 远程,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

Exocortex 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

Exocortex 需要 API Key 吗?

资料标注需要会话登录认证,使用前准备好对应凭据,并按最小权限配置。

Exocortex 支持哪些客户端?

资料标注支持 Cursor、Codex、Zed;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

Exocortex 能提供哪些能力?

资料显示能力标签包括 mcp、ai memory、knowledge graph。具体工具清单以官方文档为准。

Exocortex 需要什么运行环境?

运行时为 Cross-platform;属于远程端点,不需要在本地装运行时。

Exocortex 是本地还是远程?

它是远程端点,连服务地址即可。传输方式为HTTP · 远程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 未标注。