OPEN SOURCE · v0.9.3 ALPHA

Scattered documents go in.
Your model gets the right information.

MESA turns large, scattered document collections into lasting memory for your AI. When a question comes in, it finds what matters, picks the current version and hands it to your model as the right context.

v0.9.3 · Alpha

Ready for local evaluation, not yet for production.

Project status →

FROM DOCUMENTS TO YOUR MODEL

One memory between your archive and your model.

YOUR DOCUMENTS

  • policies/
  • contracts/2026/
  • product-notes.md
  • support-tickets/
  • meeting-notes.txt
  • regulations/
  • training/
  • faq.md

Many files, many versions, different teams.

MESA

  1. AddDocuments enter once, with their source.
  2. SplitText becomes searchable passages; model-based facts require explicit extraction.
  3. Keep versionsA change creates a new revision; the current one wins.
  4. Search four waysMeaning, words, facts and relationships, within what the app may read.
  5. Rank and limitResults are combined and trimmed to fit the model.

CONTEXT FOR YOUR MODEL

How many days do customers have to return a product?

  • “Returns are accepted within 30 days.”policies/returns.txt · v2
  • “Refunds go back to the original payment method.”support-guide.md · v5
  • “Opened software cannot be returned.”policies/returns.txt · v2

Your model writes the answer from this context.

Illustrative flow, not a live query. See a recorded real run →

THE PROBLEM

The more documents you have, the harder the right one is to find.

A language model only knows what is in front of it. A whole archive does not fit into a prompt, so something has to pick the right passages every time.

  • The whole archive does not fit.
  • Uses an outdated document.
  • Forgets what it learned yesterday.

01MEMORY

A large archive, organised

MESA keeps your documents as searchable passages, so the model gets the relevant part, not the whole archive.

Only what matters reaches the model.

See the architecture →

02REVISION

The current version wins

When a document changes, searches return the new version once it is processed. The old one stays in history.

MESA numbers every revision.

See it in the example →

03SOURCE

Every answer has a source

Your assistant can show which document and which passage supports what it says.

MESA returns the passage and its document.

How search works →

WHAT MESA DOES

Designed to be the memory of larger systems.

Many applications, agents and teams can share one memory. MESA keeps it organised, current and limited to what each of them may read.

01

Durable intake

Documents are accepted first and processed in the background. Nothing half-written reaches search.

SQLite · LanceDB · Kùzu
02

Four ways to search

Meaning, exact words, extracted facts and relationships are searched together and combined into one ranking.

vector · BM25 · assertion · graph
03

Scoped access

Each workspace, dataset and agent sees only what it is allowed to. Eligibility is checked before ranking.

tenant · workspace · dataset · agent
04

Versions and history

Every change becomes a new revision. Search uses the current one; the history stays.

revision · COMMITTED
05

Context that fits

Results are trimmed to a token budget before they reach the model, with their sources attached.

token budget · evidence
06

Connect from anywhere

The CLI, Python SDK, HTTP API and MCP all reach the same memory.

CLI · SDK · HTTP · MCP

Good to know: your own model writes the answer. MESA gives it the right text.

YOUR FIRST SUCCESSFUL SEARCH

Start small. Grow when you need to.

TodayLOCAL

Run MESA on your computer

Install the package and start the server. Word search works without any model account.

Terminal · keep this window open
pip install mesa-memory==0.9.3
mesa start

Python 3.11–3.13 · Linux / macOS · Windows with WSL or Docker

WHERE IT FITS

For organisations that run on documents.

Wherever large document collections have to feed AI reliably. These are fitting scenarios, not customer references.

Enterprise knowledge

Procedures, policies and notes from many departments, searched by assistants that only see what each team may read.

  • Procedures
  • Policies
  • Team notes

Customer support

Product documents and past tickets give support assistants the current answer, not last year’s.

  • Product docs
  • Support tickets
  • FAQ

Compliance and audit

Regulations and internal rules change; revision history shows which version an answer relied on.

  • Regulations
  • Internal rules
  • Audit records

Research and analysis

Reports and papers stay searchable by meaning and by exact terms, with the source of every finding.

  • Reports
  • Papers
  • Field notes

AI agents

Agents keep what they learned across sessions and restarts, kept apart per agent and workspace.

  • Session notes
  • Task history

BEFORE YOU TRY IT

What works, and what you need.

Text input
The CLI reads UTF-8 text files. Convert PDF and Word files to text first.
Local data
Data is stored in ~/.mesa/data. Models and external providers are off by default; this is not a full network-isolation guarantee.
Optional models
Meaning search and fact extraction need a model provider, which can receive data and charge for use.
Size limit
The document HTTP API accepts up to 5,000,000 UTF-8 bytes per request.

GET STARTED

Try it with your own documents.

$ pip install mesa-memory==0.9.3·Open source · Apache 2.0