mcp-prose-memory
This server provides persistent, sectioned memory for MCP clients through the memory and memory_context tools.
View all memories or filter by section, with formatted, compact, or JSON output.
Add facts to predefined sections like work, personal, top_of_mind, history, and instructions.
Upsert structured facts using a key/value pair, with optional confidence and source.
Remove facts by section and line number.
Replace facts by section and line number.
Load session-start memory via
memory_context, optionally bounded by selected sections andmaxChars.Persist facts across sessions in a JSON memory file with atomic writes.
Enforce limits such as 300 characters per fact and 80 characters per structured key.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mcp-prose-memory add that my favorite color is blue"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
mcp-prose-memory
Persistent memory for MCP clients.
mcp-prose-memory gives an agent a small, durable place to keep facts across sessions. It stores memory as JSON, keeps facts grouped by section, and exposes one tool for careful add, remove, replace, and view operations.
It is built for memory that should survive restarts without becoming a loose text file that slowly drifts out of shape.
Docs · Changelog · Contributing · Security
Features
JSON memory storage with a stable schema
Compact structured facts with optional
key,value,source, andconfidenceAtomic writes through temp-file replacement
Atomic fact operations: add, remove, replace, view
Sectioned context organization for general clients and local assistant memory
Case-insensitive duplicate detection
Strict line-number validation for remove and replace
Automatic normalization for older or partial JSON documents
Compact
memory_contextoutput with section and character-budget filtersLimits: 30 facts per section, 300 characters per fact, 80 characters per structured key
Configurable storage path via environment variable
Related MCP server: agentsmd-memory
Installation
Full setup guide: Docs.
Runtime: Node >= 18. Node 20 or newer is recommended, since Node 18 is end-of-life. On Node 18, npm prints an EBADENGINE warning for @hono/node-server, a dependency of the MCP SDK's HTTP transport that this stdio server never loads. The warning is harmless, but installs with engine-strict=true (or Yarn 1) need Node 20+ or engine checks turned off.
npm install -g mcp-prose-memoryOr run with npx:
npx mcp-prose-memoryConfiguration
Configuration reference: Docs.
Default storage is ~/.mcp-prose-memory/memory.json. Override it with MEMORY_PATH.
If you used an older release with a client-specific default memory location, either move that JSON file to the new default path or set MEMORY_PATH to the existing file.
Desktop Client
Add a server entry like this to your MCP client config:
{
"mcpServers": {
"memory": {
"command": "npx",
"args": ["mcp-prose-memory"]
}
}
}CLI Client
Add a server entry like this to your CLI MCP config:
{
"mcpServers": {
"memory": {
"command": "npx",
"args": ["mcp-prose-memory"]
}
}
}Custom Memory Location
{
"mcpServers": {
"memory": {
"command": "npx",
"args": ["mcp-prose-memory"],
"env": {
"MEMORY_PATH": "/path/to/your/memory.json"
}
}
}
}Memory File
The memory file is JSON with arrays of facts per section:
{
"version": 5,
"updated": "2025-01-15T10:30:00.000Z",
"sections": {
"work": ["Fact 1", "Fact 2"],
"personal": ["Lives in Berlin", "Prefers dark mode"],
"top_of_mind": [],
"history": ["Completed project X"],
"instructions": ["Be concise"],
"user_preferences": [
{
"key": "answer_style",
"value": "Prefers concise answers",
"confidence": "high",
"source": "user_explicit",
"createdAt": "2025-01-15T10:30:00.000Z",
"updatedAt": "2025-01-15T10:30:00.000Z"
}
]
}
}If the file does not exist, the server starts with an empty document. If the file is invalid JSON, the server fails the operation instead of wiping memory.
Tools
Tools reference: Docs.
memory
Single tool for all memory operations. The command parameter selects the action.
Commands:
view
Show all memories or filter by section.
{"command": "view"}
{"command": "view", "section": "work"}add
Add a fact to a section.
{"command": "add", "section": "personal", "fact": "Lives in Berlin"}Structured compact facts are also supported:
{"command": "add", "section": "user_preferences", "key": "answer_style", "value": "Prefers concise answers"}upsert
Add a structured fact or replace the existing fact with the same key.
{"command": "upsert", "section": "user_preferences", "key": "answer_style", "value": "Prefers concise and direct answers"}remove
Remove a fact by line number.
{"command": "remove", "section": "work", "line": 3}replace
Update a fact by line number.
{"command": "replace", "section": "top_of_mind", "line": 1, "fact": "Working on new project"}memory_context
Returns memory for session initialization. Clients can request compact bounded context.
{}{"format": "compact", "sections": ["user_profile", "user_preferences"], "maxChars": 1500}Sections
Section | Purpose |
| Professional context, projects, colleagues, tools |
| Location, preferences, interests, personal facts |
| Current focus, active tasks |
| Past events, completed work |
| Standing rules, behavioral preferences |
| Stable user profile facts |
| Durable preferences |
| Eyra-specific product and architecture facts |
| Durable local environment facts |
| Repeated workflow preferences |
| Writing and tone preferences |
| Durable task context |
| Explicitly requested durable reminders |
Development
git clone https://github.com/gabrimatic/mcp-prose-memory.git
cd mcp-prose-memory
npm ci
npm test
npm run checknpm test builds the TypeScript source and runs store-level plus real MCP stdio smoke tests. npm run check also runs a production dependency audit and verifies the npm package contents with npm pack --dry-run.
prepublishOnly runs the same check before publishing.
Project Support
Changes: CHANGELOG.md
Contributions: CONTRIBUTING.md
Security reports: SECURITY.md
Issues: GitHub Issues
Developer
By Soroush
License
MIT
Available Tools
2 toolsmemoryA
Manage persistent memory across sessions.
Commands:
view: Show all memories (optionally filter by section)
add: Add a new fact to a section
upsert: Add or replace a structured fact by key
remove: Remove a fact by section and line number
replace: Update a fact by section and line number
Sections: work, personal, top_of_mind, history, instructions, user_profile, user_preferences, eyra_project, devices_environment, workflows, writing_style, long_term_tasks, do_not_forget
Examples:
{"command": "view"}
{"command": "view", "section": "work"}
{"command": "add", "section": "personal", "fact": "Lives in Berlin"}
{"command": "upsert", "section": "user_preferences", "key": "answer_style", "value": "Prefers concise answers"}
{"command": "remove", "section": "work", "line": 3}
{"command": "replace", "section": "top_of_mind", "line": 1, "fact": "Working on new project"}
Inputs: command (required), key, fact, line, +7 more.
memory_contextARead-only
Load memory for session start. Supports compact bounded context for clients with small prompt budgets.
Inputs: format, maxChars, sections.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 2 tools
The two tools have clearly distinct purposes: 'memory' handles CRUD operations on memories, while 'memory_context' loads memory for session start. No overlap or ambiguity.
Both tool names start with 'memory' and use underscores for compound names, following a consistent pattern. However, 'memory' is a single noun while 'memory_context' is a compound noun, a minor deviation.
With only 2 tools, the server is lean. All CRUD operations are bundled into one tool, which works but limits granularity. The count is on the low side but acceptable given the internal richness.
The 'memory' tool covers all standard CRUD operations (view, add, upsert, remove, replace), and 'memory_context' handles context loading. Minor gaps like listing sections or clearing all memories are missing but not critical.
Maintenance
Related MCP Connectors
Hosted agent memory: store, search, and recall facts across sessions from any MCP client.
Hosted memory for AI agents over MCP. One memory shared by all your agents.
An MCP memory server. One memory your agents share — across models, devices and apps.
Cloud-hosted MCP server for durable AI memory
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceMCP server providing persistent memory for LLMs via semantic search over captures, structured facts with temporal tracking, and document ingestion. Enables any MCP-compatible client to maintain a searchable knowledge base across sessions.ISC
- AlicenseAqualityBmaintenanceMCP server for managing project memory in AGENTS.md. Allows agents to record and remove durable facts using save/forget tools.338 npm1MIT
- FlicenseNot gradedqualityCmaintenancePersistent memory MCP server for AI agents that stores, recalls, and searches conversation history, key-value context, and long-term entries across sessions with semantic search and FIFO queues.110 npm1-
- AlicenseAqualityCmaintenanceAn MCP server that gives agents persistent memory with namespaced facts, tags, full-text search, and TTL expiry, all running locally on SQLite with zero external dependencies.83MIT