It grows without bound
Every new convention, edge case, and workaround gets appended. What started as a helpful guide becomes an unreadable wall of text.
Beta — Team memory with admin panel
AI memory your whole team can trust together
mex gives engineering teams a shared, structured memory system for AI coding agents — with an admin panel to keep context aligned, validated, and current as your project grows.
Looking for the local open-source version? View on GitHubThe open-source version of mex runs locally on your machine. View the source and documentation on GitHub.
The problem
The hidden cost of giant instruction files
Most teams start with a single rules or instructions file per developer. It works at first — then quietly fragments collaboration and degrades AI output across the team.
It goes stale quickly
The codebase evolves daily. Your instructions file does not. Agents end up following rules that no longer match reality.
Teams drift apart
Each developer maintains their own version. Context becomes inconsistent across machines, repos, and team members.
Tokens add up fast
Sending the entire file with every request burns context window and budget — paying for outdated information repeatedly.
Output quality degrades
Over time, agents lose focus. Suggestions become generic, miss project nuance, and require more correction from you.
The solution
Team memory with an admin panel built in
mex replaces scattered instructions files with a shared memory system your whole team works from. An admin panel gives leads visibility and control — so context stays organized, validated, and aligned as people, priorities, and code change.
The result: every developer's AI agents draw from the same trusted source, and your team spends less time correcting drift and more time shipping.
- Shared memory that keeps every teammate and agent on the same page
- Admin panel to manage, review, and govern team context in one place
- Continuous validation so memory stays accurate as the codebase evolves
- Agents that remain reliable across weeks and months of team development
How it works
Shared memory your whole team can manage
Four steps from scattered context to team-wide reliability — with an admin panel at the center.
Organize team context
Structure project knowledge into shared layers — conventions, architecture, priorities — instead of one file per developer.
Keep memory validated
mex continuously checks memory against your codebase so every teammate's agents work from context that is still true.
Manage from the admin panel
Review, update, and govern shared memory in one place. Team leads see what agents know and fix drift before it spreads.
Stay aligned over time
As your team and codebase grow, mex keeps everyone working from the same trusted memory — week after week.
Who it's for
Built for teams that work on AI-assisted code together
mex is designed for collaborative development — with an admin panel so leads can keep shared context honest and current.
Small engineering teams
- One shared memory layer instead of conflicting rules files
- Admin panel to see and manage what agents know
- Consistent AI output across every developer on the team
Team leads & admins
- Central visibility into project memory and context health
- Govern who can update conventions, decisions, and priorities
- Catch drift before it spreads across the team
Scaling organizations
- Memory infrastructure that survives growth and turnover
- Admin controls for consistency across repos and squads
- Reliable AI workflows at engineering-org scale
Why we built mex
Our team kept fixing the same AI mistakes
We started like most teams — everyone maintaining their own instructions file, agents drifting apart within weeks. Corrections piled up, context went stale, and nobody had a clear view of what our AI tools actually knew. We built mex so teams can share structured, validated memory — with an admin panel to keep it honest — instead of hoping individual rules files stay in sync. AI-assisted development should get better as your team grows, not harder.
MIT Licensed · Open Source
Team beta now. Open source runs locally.
The team version of mex — with shared memory and an admin panel — is available through our beta program. Request access for your team above.
The open-source version of mex runs locally on your machine. View the source and documentation on GitHub. It is MIT licensed and maintained at mex-memory/mex on GitHub.
Insights
Thinking about AI memory
Perspectives on context drift, team alignment, and building reliable AI workflows that last.
2 min read
Structured memory is the missing layer in AI-assisted development
Agents, models, and IDEs get most of the attention. But without memory infrastructure, none of it scales past the first month.
Read more2 min read
The real cost of context drift in growing teams
When every developer has a slightly different version of the truth, AI output becomes inconsistent — and trust erodes fast.
Read more2 min read
Why giant context files quietly destroy long-term AI coding performance
The instructions file that worked on day one becomes a liability by week four. Here is why — and what actually breaks.
Read moreGet beta access for your team
Shared memory, an admin panel, and agents that stay aligned — built for teams working on code together.