Generated banner for MIRIX
Generated logo for MIRIX

MIRIX

Memory server

Created by Mirix AI

Published in Mirix-AI/MIRIX. Publisher page

3,460 GitHub stars

What it is

Mirix is a multi-agent personal assistant designed to track on-screen activities and answer user questions intelligently. By capturing real-time visual data and consolidating it into structured memories, Mirix transforms raw inputs into a rich knowledge base that adapts to your digital experiences.
— the description by Mirix AI

AutoGineer overview, from the brain’s own files

Mirix is a memory server designed to provide AI agents with persistent, multi-modal memory. It functions by observing screen activity and processing natural language conversations, consolidating these inputs into six distinct memory components: Core, Episodic, Semantic, Procedural, Resource, and Knowledge Vault. The system uses a hybrid storage approach, leveraging PostgreSQL with pgvector for search and an optional Neo4j graph database for provenance tracking. Mirix includes an 'Auto-Dream' feature for periodic memory consolidation, deduplication, and refinement, ensuring that the agent's knowledge base remains structured and relevant over time.

How it works

Memory model: Hybrid

AutoGineer overview, from the brain’s own files

Mirix operates as a backend server that orchestrates six specialized sub-agents to manage different memory types. It captures user interactions and visual screen data, which are then processed and stored in a PostgreSQL database. The system supports full-text search via BM25 and vector similarity search. An optional graph memory layer can be enabled to map relationships between entities using Neo4j. The Auto-Dream functionality allows the system to review and merge memory entries, while the provided Python client enables agents to read from and write to these memory stores.

Folders
FolderRoleFiles
.githubagent-config1
assetsattachments3
docsdocs5
samplesexamples16
scriptsscripts34
teststests132
.vscodefolder2
dashboardfolder35
dockerfolder3
evalsfolder53
mirixfolder341

The agent writes

Requirements

Apps: Docker

Runtimes: Python >=3.10,<4.0

Databases: Neo4j, PostgreSQL, Redis, SQLite

API keys (by name): MIRIX_API_KEY

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Google Antigravity, Kiro, OpenCode, Amp, Roo Code, Cline, JetBrains Junie, Devin (ex-Windsurf)

Set it up

  1. This brain runs code automatically: 2 scripts. Review Makefile, setup.py before opening the folder in Claude Code. No step below turns a hook on for you.

    • Makefile: $(POETRY) install
    • setup.py: python setup.py
  2. Not scanned by AutoGineerThe author’s README also describes installing it with Docker Compose. Read its install section: AutoGineer did not scan that path. https://github.com/Mirix-AI/MIRIX/blob/9077e4b229dd8f61c5703b431a97788c6b61acfb/README.md#quick-start
  3. Not scanned by AutoGineerAutoGineer could not scan everything it runs. https://pypi.org/project/mirix/
  4. Created by Mirix AI. Published in Mirix-AI/MIRIX. Licence: Apache License 2.0.

Usage

AutoGineer overview, from the brain’s own files

  • Initialize a meta-agent with specific memory components.
  • Add conversation turns or tool-use logs to the memory store.
  • Retrieve context-aware memories based on user queries.
  • Trigger memory consolidation using the Auto-Dream endpoint.
  • Enable graph-based provenance for complex entity relationships.
“What did we discuss on MirixDB in last 4 days?”
The agent retrieves relevant conversation history and memory blocks from the last four days to answer the user's question.
“Deploy the service.”
The agent records the tool call, captures any resulting errors or successful fixes, and stores this as procedural memory for future reference.

Documentation by Mirix AI

Image: Mirix Logo

MIRIX - Multi-Agent Personal Assistant with an Advanced Memory System

Your personal AI that builds memory through screen observation and natural conversation

| 🌐 Website | 📚 Documentation | 📄 Paper | 💬 Discord

<!-- | [Twitter/X](https://twitter.com/mirix_ai) | [Discord](https://discord.gg/S6CeHNrJ) | -->


Key Features 🔥
  • Multi-Agent Memory System: Six specialized memory components (Core, Episodic, Semantic, Procedural, Resource, Knowledge Vault) managed by dedicated agents

  • Screen Activity Tracking: Continuous visual data capture and intelligent consolidation into structured memories

  • Privacy-First Design: All long-term data stored locally with user-controlled privacy settings

  • Advanced Search: PostgreSQL-native BM25 full-text search with vector similarity support

  • Graph Memory (optional): A provenance hypergraph over episodic and semantic memories in Neo4j, with n-ary facts that link entities through named roles and cite the memory rows they came from

  • Multi-Modal Input: Text, images, voice, and screen captures processed seamlessly

Results

Image: MIRIX-2.0 compared with other memory systems on user-centric and agent-centric tasks

MIRIX-2.0 compared with other memory systems. Each axis label gives the MIRIX-2.0 score; the value in parentheses is its lead over the strongest baseline on that axis, shown where MIRIX-2.0 leads.

  • (a) User-centric tasks (accuracy, %): conversational recall on LoCoMo and LongMemEval-S, and the document QA tracks of MemoryAgentBench (SHDocQA, MHDocQA, DetectiveQA). Baselines are A-Mem, Graphiti, MemoryBank, mem0 and HyperMem. LoCoMo uses graph memory with AutoDream; LongMemEval-S and document QA use graph memory alone.

  • (b) Agent-centric tasks (success rate, %): the six ALFWorld task types (Pick, Look, Clean, Heat, Cool, Pick Two) and five WebShop product categories (Fashion, Beauty, Electronics, Garden, Grocery), with GPT-6 Luna as the agent model. Baselines are no memory, AWM and ReasoningBank.

Quick Start

Step 1: Backend & Dashboard (Docker):

docker compose up -d --pull always

Step 2: Create an API key in the dashboard (http://localhost:5173) and set as the environmental variable MIRIX_API_KEY.

Step 3: Client (Python, mirix-client, https://pypi.org/project/mirix-client/):

pip install mirix-client

Now you are ready to go! See the example below:

from mirix import MirixClient

client = MirixClient(
    api_key="your-api-key",
    base_url="http://localhost:8531",
)

client.initialize_meta_agent(
    config={
        "llm_config": {
            "model": "gemini-3.6-flash",
            "model_endpoint_type": "google_ai",
            "api_key": "your-api-key-here",
            "model_endpoint": "https://generativelanguage.googleapis.com",
            "context_window": 1_000_000,
        },
        "embedding_config": {
            "embedding_model": "gemini-embedding-001",
            "embedding_endpoint_type": "google_ai",
            "api_key": "your-api-key-here",
            "embedding_endpoint": "https://generativelanguage.googleapis.com",
            "embedding_dim": 768,
        },
        "meta_agent_config": {
            "agents": [
                {
                    "core_memory_agent": {
                        "blocks": [
                            {"label": "human", "value": ""},
                            {"label": "persona", "value": "I am a helpful assistant."},
                        ]
                    }
                },
                "resource_memory_agent",
                "semantic_memory_agent",
                "episodic_memory_agent",
                "procedural_memory_agent",
                "knowledge_vault_memory_agent",
            ],
        },
    }
)

client.add(
    user_id="demo-user",
    messages=[
        {"role": "user", "content": [{"type": "text", "text": "The moon now has a president."}]},
        {"role": "assistant", "content": [{"type": "text", "text": "Noted."}]},
    ],
    session_id="sess-demo-001",
)

# Include tool activity when you want skills learned from the WORK PROCESS:
# tool errors, retries, and the fix that finally worked are the distiller's
# strongest signals. Pass tool results as role="tool" turns and/or a
# "tool_calls" list on assistant messages (OpenAI shape) — both are preserved
# in the session conversation store.
client.add(
    user_id="demo-user",
    messages=[
        {"role": "user", "content": [{"type": "text", "text": "Deploy the service."}]},
        {"role": "assistant", "content": "", "tool_calls": [
            {"function": {"name": "kubectl_apply", "arguments": "{\"file\": \"deploy.yaml\"}"}}
        ]},
        {"role": "tool", "name": "kubectl_apply", "content": "error: forbidden (missing RBAC)"},
        {"role": "assistant", "content": [{"type": "text", "text": "Fixed the RBAC role and redeployed successfully."}]},
    ],
    session_id="sess-demo-001",
)

memories = client.retrieve_with_conversation(
    user_id="demo-user",
    messages=[
        {"role": "user", "content": [{"type": "text", "text": "What did we discuss on MirixDB in last 4 days?"}]},
    ],
    limit=5,
)
print(memories)

For more API examples, see samples/run_client.py.

Auto-Dream Memory Consolidation

MIRIX also exposes an auto-dream endpoint for explicit memory cleanup and consolidation. Auto-dream reviews existing memories, merges duplicates, resolves stale/conflicting entries where possible, and writes the result back through the memory tools.

REST endpoint:

POST /memory/auto_dream?user_id=demo-user
Content-Type: application/json

{
  "mode": "experience",
  "dry_run": false
}

Supported mode values:

ModeMemory component(s) reviewed
coreCore memory blocks
episodicEpisodic memories
semanticSemantic memories
resourceResource memories
proceduralProcedural memories
knowledgeKnowledge Vault memories
experienceEpisodic, Semantic, and Knowledge Vault memories together

Use dry_run: true to fetch counts and inspect what would be processed without applying updates. The optional model field can override the auto-dream agent model for that request. For procedural mode, pass the target meta_agent_id and optionally last_n_sessions to distill recent session experiences.

Python client example:

result = client.auto_dream(
    user_id="demo-user",
    mode="procedural",
    meta_agent_id=meta_agent.id,
    last_n_sessions=3,
    dry_run=False,
)
print(result)
Graph Memory

MIRIX can index episodic and semantic memories into a provenance hypergraph stored in Neo4j. Entities become anchors, each extracted fact becomes an n-ary frame whose edges name the role each anchor plays, and every frame cites the memory rows it came from. With graph memory on, episodic and semantic retrieval enters through the graph, and Auto-Dream merges duplicate anchors on the graph instead of rewriting stored memories.

Graph memory is off by default. To turn it on:

  1. Start Neo4j 5.23 or later. docker compose --profile graph up -d starts Neo4j 5.26 with APOC; first set MIRIX_NEO4J_PASSWORD (at least 8 characters) in .env, because there is no default password.

  2. Install the Neo4j driver. It is in requirements.txt; after pip install -e ., run pip install "neo4j>=5.20,<6".

  3. Set MIRIX_ENABLE_GRAPH_MEMORY=true and MIRIX_NEO4J_PASSWORD to the password your Neo4j uses. MIRIX_NEO4J_URI and MIRIX_NEO4J_USER default to bolt://localhost:7687 and neo4j.

With graph memory on, POST /memory/auto_dream refines the graph only unless the request sets graph_only to false. Evaluation settings and commands are in evals/GRAPH_MEMORY.md.

Upgrading an Existing Database

Fresh databases get their full schema automatically at startup. Existing databases require manual migrations when upgrading to a version that alters tables — startup creates missing tables but never adds columns to existing ones. If migrations are pending, the server refuses to start and names the exact scripts to run.

For the skill-based procedural memory release, run against your database (in this order):

psql "$MIRIX_PG_URI" -f scripts/migrate_procedural_to_skill.sql
psql "$MIRIX_PG_URI" -f scripts/migrate_add_message_session_id.sql
psql "$MIRIX_PG_URI" -f scripts/migrate_add_message_session_id_phase2.sql   # outside a transaction (CREATE INDEX CONCURRENTLY)
psql "$MIRIX_PG_URI" -f scripts/migrate_add_conversation_message.sql
psql "$MIRIX_PG_URI" -f scripts/migrate_add_conversation_message_phase2.sql # outside a transaction
psql "$MIRIX_PG_URI" -f scripts/migrate_add_skill_experience.sql
psql "$MIRIX_PG_URI" -f scripts/migrate_add_agent_trigger_state.sql
psql "$MIRIX_PG_URI" -f scripts/migrate_backfill_procedural_scope.sql   # backfill filter_tags['scope'] from each skill's owning client

For the graph-memory release, which adds five provenance columns, also run:

psql "$MIRIX_PG_URI" -f scripts/migrate_add_graph_provenance_columns.sql

Each script is idempotent where possible and documents its own preconditions in its header. Run them once, then restart the server.

License

Mirix is released under the Apache License 2.0. See the LICENSE file for more details.

Contact

For questions, suggestions, or issues, please open an issue on the GitHub repository or contact us at [e-mail address on the upstream page](https://github.com/Mirix-AI/MIRIX/blob/9077e4b229dd8f61c5703b431a97788c6b61acfb/README.md)

Join Our Community

Connect with other Mirix users, share your thoughts, and get support:

💬 Discord Community

Join our Discord server for real-time discussions, support, and community updates: https://discord.gg/S6CeHNrJ

🎯 Weekly Discussion Sessions

We host weekly discussion sessions where you can:

  • Discuss issues and bugs

  • Share ideas about future directions

  • Get general consultations and support

  • Connect with the development team and community

📅 Schedule: Friday nights, 8-9 PM PST 🔗 Zoom Link: https://ucsd.zoom.us/j/96278791276

📱 WeChat Group

You can add the account ari_asm so that I can add you to the group chat.

Acknowledgement

We would like to thank Letta for open-sourcing their framework, which served as the foundation for the memory system in this project.

CONTRIBUTING.md

Contributing to MIRIX

We’re thrilled to have you join us in contributing to MIRIX! As passionate champions of open-source collaboration, we’re committed to building not only a powerful tool but also a vibrant community where developers of all skill levels can make impactful contributions. Before you get started, we encourage you to review the MIRIX Code Standards. This will help ensure consistency and set the stage for success for everyone involved.

Our contribution pathways are designed to make it easy for you to get involved and start making an impact immediately:

Four Ways to Get Involved

Help With Existing Issues

Our developers frequently tag issues with "help wanted" and "good first issue," highlighting pre-vetted tasks with a clear scope. Plus, there's always someone available to assist you if you encounter any challenges along the way.

Create Your Own Tickets

Noticed something that needs fixing? Got an idea for an improvement? You don’t need permission to identify problems—the people closest to the pain are often the best equipped to envision solutions.

For feature requests, share the story of what you’re trying to achieve. What are you working on? What’s standing in your way? What would make your life easier? Submit your ideas through our GitHub issue tracker using the "Feature Request" label.

For bug reports, it’s important to provide enough details for us to reproduce the issue. Use our GitHub issue tracker and include the following:

  • A clear and concise title summarizing the problem

  • What you were attempting to do when the issue occured

  • What you expected to happen versus what actually happened

  • A code sample, test case, or detailed steps to help us replicate the issue

Share Your Use Cases

Not all contributions are about writing code—sometimes the most impactful ones come from sharing how you're using our project. If you're leveraging MIRIX in an interesting way, consider adding it to the samples folder. This not only helps others explore new possibilities but also counts as a meaningful contribution. We often highlight compelling examples in our blog posts and videos, so your work could be showcased to the broader community!

Help Others in Discord

Join our Discord community and contribute by lending a hand at the helpdesk. Answering questions and helping troubleshoot issues is an invaluable way to support the community. The knowledge you share today could save someone hours of frustration tomorrow.

What happens next?

Notes for Large Changes

Please keep changes as concise as possible. For major architectural changes (>500 LOC), please create a GitHub issue (RFC) to discuss the technical design and justification. Otherwise, we may tag it as "rfc-required" and potentially not review the PR.

Once you've identified an issue tagged with "good first issue" or "help wanted," or have an example you'd like to share, here's how to turn it into a contribution:

  1. Share your proposed approach in the issue discussion or on Discord before diving into the code. This ensures your solution aligns with MIRIX's architecture from the outset and helps avoid potential rework.

  2. Fork the repository, create a branch for your changes, and submit a pull request (PR). Detailed technical instructions are provided below — be open to feedback during the review process.

Setup

  1. Fork the repository on GitHub.

  2. Clone your fork locally:

    git clone https://github.com/Mirix-AI/MIRIX
    cd MIRIX
  3. Set up your development environment:

    Follow the "Development Installation" section in this Get Started document.

    If you prefer to poetry instead of pip, follow these additional steps. (uv follows similar steps)

    • Ensure you have Python 3.11+ installed.

    • Install poetry: https://python-poetry.org/docs/

    • [Recommended] Configure poetry to create virtual env under your current project

      poetry config virtualenvs.in-project true
    • Install project dependencies:

      make install

Making Changes

  1. Create a new branch for your changes:

    git checkout -b your-branch-name
  2. Make your changes in the codebase.

  3. Write or update tests as necessary.

  4. Run the tests to ensure they pass:

    make test
  5. Format your code:

    make format
  6. Run linting checks:

    make lint

Submitting Changes

  1. Commit your changes:

    git commit -m "Your commit message"
  2. Push to your fork:

    git push origin your-branch-name
  3. Submit a pull request through the GitHub website to https://github.com/Mirix-AI/MIRIX.

Pull Request Guidelines

  • Provide a clear title and description of your changes.

  • Include any relevant issue numbers in the PR description.

  • Ensure all tests pass and there are no linting errors.

  • Update documentation if you're changing functionality.

Code Style and Quality

We use several tools to maintain code quality:

  • Ruff for linting and formatting

  • Pyright for static type checking

  • Pytest for testing

Before submitting a pull request, please run:

make check

This command will format your code, run linting checks, and execute tests.

Third-Party Integrations

When contributing integrations for third-party services (LLM providers, embedding services, databases, etc.), please follow these patterns:

Optional Dependencies

All third-party integrations must be optional dependencies to keep the core library lightweight. Follow this pattern:

  1. Add to pyproject.toml: Define your dependency as an optional extra AND include it in the dev extra:

    [project.optional-dependencies]
    your-service = ["your-package>=1.0.0"]
    dev = [
        # ... existing dev dependencies
        "your-package>=1.0.0",  # Include all optional extras here
        # ... other dependencies
    ]
  2. Use TYPE_CHECKING pattern: In your integration module, import dependencies conditionally:

    from typing import TYPE_CHECKING
    
    if TYPE_CHECKING:
        import your_package
        from your_package import SomeType
    else:
        try:
            import your_package
            from your_package import SomeType
        except ImportError:
            raise ImportError(
                'your-package is required for YourServiceClient. '
                'Install it with: pip install MIRIX[your-service]'
            ) from None
  3. Benefits of this pattern:

    • Fast startup times (no import overhead during type checking)

    • Clear error messages with installation instructions

    • Proper type hints for development

    • Consistent user experience

  4. Do NOT:

    • Add optional imports to __init__.py files

    • Use direct imports without error handling

    • Include optional dependencies in the main dependencies list

Integration Structure
  • Place LLM clients in mirix/llm_client/

  • Place embedding clients in mirix/embedder/

  • Place database drivers in mirix/driver/

  • Place database script python files in database/

  • Place samples in samples/

  • Follow existing naming conventions (e.g., your_service_client.py)

Testing
  • Add comprehensive tests in the appropriate tests/ subdirectory

  • Mark integration tests with _int suffix if they require external services

  • Include both unit tests and integration tests where applicable

Questions?

Stuck on a contribution or have a half-formed idea? Come say hello in our Discord server. Whether you're ready to contribute or just want to learn more, we're happy to have you! It's faster than GitHub issues and you'll find both maintainers and fellow contributors ready to help.

Thank you for your contribution to MIRIX!

512 files · 499 not served here

Advantages

  • Supports multi-modal inputs including text, images, voice, and screen captures. evidence: README.md
  • Provides a provenance hypergraph to link entities and cite memory sources. evidence: README.md

Limitations

  • Existing databases require manual migration scripts when upgrading to versions that alter table schemas. evidence: README.md
  • Requires manual configuration and setup of external dependencies like PostgreSQL and optional Neo4j. evidence: README.md

Security

Risk level: Low risk

This brain runs

Agent instructions that run or fetch something (12)

  • .cursorrules: User-defined tools run in `ToolExecutionSandbox` for security
  • .cursorrules: Run tests with: `pytest tests/`
  • .cursorrules: Run linters before committing: `ruff check .` and `pyright`
  • .cursorrules: User-defined tools MUST run in `ToolExecutionSandbox`
  • CLAUDE.md: docker-compose up -d # starts PostgreSQL (5432), Redis (6379), API (8531), Dashboard (5173)
  • CLAUDE.md: docker-compose ps # verify all services healthy
  • CLAUDE.md: pip install -e .
  • CLAUDE.md: python scripts/start_server.py --port 8531
  • CLAUDE.md: python scripts/start_server.py --port 8899 # Terminal 1
  • CLAUDE.md: make format # ruff import sort + format
  • CLAUDE.md: make lint # ruff check + pyright
  • CLAUDE.md: make check # format + lint + test

Install scripts (2)

  • Makefile (make install): $(POETRY) install
  • setup.py: python setup.py
  • Low risk: network-instructions (CLAUDE.md:31)
  • Low risk: scripts-present (scripts/start_server.py)
  • Low risk: scripts-present (setup.py)

Attribution and licence

Created by
Mirix AI
First committed by
@wangyu-ustc
Declared authors
  • Mirix AI (declared in pyproject.toml)
Publisher
MIRIX-AI (Community)
Source
Mirix-AI/MIRIX · commit 9077e4b
Licence
Apache License 2.0 (files served here)
  • ATTRIBUTION.md (written by AutoGineer)
Synced

AutoGineer is not affiliated with or endorsed by Mirix AI.

Install from source

499 of its files are not served here, so there is no ZIP. Mirix-AI/MIRIX at 9077e4b