TalvikAI infrastructure for portable, governed context.
Talvik is an AI infrastructure company building the open foundation for context that stays reusable as teams move across AI tools, models and agents. MAK4I is our open protocol for it — a working reference implementation is running in Developer Preview.
The protocol for not rebuilding what your AI already built.
- MIT licensed
- Independently implementable
- Portable context
- Working reference implementation
Current Status
Developer Preview
A working reference implementation already covers the core protocol workflow. Here's exactly where each part stands.
- Protocol specification
- Ongoing
- Defines portable context structures, lifecycle behavior, provenance, versioning and interoperability requirements; refined alongside the implementation.
- Reference implementation
- Hosted & running
- A hosted MAK4I reference implementation is running as an access-controlled Developer Preview, validated against the evolving specification.
- Context lifecycle
- Working
- Create, retrieve, search and supersede — with version history, lineage and provenance on every artifact.
- Authorization boundaries
- Implemented
- Organization and project authorization boundaries are enforced, with individual principals and credentials.
- MCP integration
- Working in the preview
- MAK4I can be reached through MCP in the reference implementation today. Broader interoperability work is ongoing.
- Developer validation
- Early
- External developers, AI implementation teams and early design partners are beginning to test against the hosted preview.
What Talvik Is
An AI infrastructure company
Talvik builds the open foundation that keeps project context and knowledge — architecture decisions, requirements, research, specifications, prompts, workflows, constraints, reusable artifacts and their history — usable as teams move across AI tools, models and agents.
MAK4I — the open protocol — sits at the center of that work. Talvik develops it in the open and builds the commercial layer around it: managed hosting, governance, administration, observability and enterprise integrations.
The Problem
Knowledge doesn't travel with you
Organizations run more AI tools, models and agents every quarter. Project knowledge fragments across all of them, and the same work gets rebuilt again and again.
Every AI tool stores knowledge differently, with no shared standard.
Context gets trapped inside whichever platform created it.
Switching tools means copying, reconstructing or regenerating context that already exists.
Teams regenerate architecture decisions, specs and workflows instead of reusing them.
The AI tool changes often. The organization's knowledge shouldn't have to start over.
Why Now?
Fragmentation is still the default
AI adoption is accelerating, but knowledge stays fragmented across tools. Teams recreate context, decisions, workflows and artifacts because each platform generates and retains them differently. MAK4I is a tool-neutral protocol for governed reuse — and a working implementation already runs to prove the approach.
Meet MAK4I
More than AI memory
MAK4I standardizes how reusable context, artifacts and knowledge are described, discovered, versioned and governed — so AI systems retrieve the current authoritative context, with lineage and provenance behind it. Memory is one artifact type, not the whole protocol.
Core Principle
Check Before Create
MAK4I promotes reuse before regeneration — search first, adapt when possible, and create only as a last resort.
- 1
Need something
- 2
Search existing artifacts
- 3
Reuse
- 4
Adapt
- 5
Only create when nothing suitable exists
What Talvik Builds — Commercial Direction
Enterprise infrastructure around the protocol
Available today: MAK4I and a hosted reference implementation you can try. Planned: the enterprise capabilities below, for organizations that need governed AI context at scale — in development, not generally available.
Managed MAK4I infrastructure
A hosted runtime for teams that don't want to operate the protocol themselves.
Enterprise context governance
Policy, access control and approval workflows over shared AI context.
Policy & administration
Central controls for how context is created, reused and retired across an organization.
Observability & audit
Visibility into what context was retrieved, by which client, and why.
Enterprise integrations & storage
Customer-controlled storage or Talvik-managed options, plus integration with existing systems.
Private & hybrid deployment
Private, hybrid and self-managed deployment support, plus implementation and support services.
No product revenue, production integrations or enterprise deployments exist today. This is the intended commercial direction around the open protocol, not a list of shipping products.
How It Works
Open protocol. Commercial infrastructure.
The protocol and the company are deliberately separate.
Open protocol
MAK4I
An open, tool-neutral interoperability protocol defining how AI context is represented, exchanged, versioned, governed and reused across systems. MIT licensed, independently implementable, with a working reference implementation — Talvik's implementation is not the definition of the protocol.
AI infrastructure company
Talvik
The company developing enterprise infrastructure, managed services and tooling around MAK4I — in development, not yet generally available — and the current steward of the protocol. The protocol never depends on buying Talvik's services.
High-Level Architecture
Deployment-neutral and storage-independent
A common layer that AI tools read from and write to. MAK4I does not require any specific cloud, orchestration platform, database or storage technology.
AI Tools & Agents
MAK4I Protocol
Represents, governs, versions and resolves reusable AI context on behalf of connected AI tools.
Portable AI Context
Decisions, architecture, workflows, history and more — in a standard, portable form.
Storage adapters
Where context is persisted — pluggable and storage-independent, not tied to any database or vendor.
Potential deployment models
- Talvik-hosted runtime with customer-controlled storage
- Fully Talvik-managed cloud deployment
- Customer-managed cloud deployment
- Private / on-premises deployment
- Hybrid deployment
- Fully self-managed MAK4I implementation
These are architectural and commercial deployment models, not a list of currently available products. The environments above are examples MAK4I has been exercised against — not official partnerships or endorsements; see the roadmap for what is built versus planned.
Where MAK4I Fits
MCP connects tools. MAK4I connects their context.
Agent-to-agent protocols connect agents. MAK4I connects their context and knowledge — an interoperability layer that sits alongside connectivity standards rather than competing with them.
MCP
What tools and data can the AI access? Connects AI systems to tools and data.
Agent-to-agent protocols
Which agents can work together? Connect agents to other agents.
MAK4I
What does the organization already know that these AI systems should be able to reuse? Provides a portable, governed context layer across them.
The MAK4I reference implementation uses MCP as one integration path. The protocol itself stays tool-neutral and does not depend on MCP.
Roadmap Preview
Where MAK4I is headed
- 1
Phase 0 — Foundation
- 2
Phase 1 — Protocol Specification
- 3
Phase 2 — Reference Implementation
- 4
Phase 3 — Reusable Artifacts
- 5
Phase 4 — Hosted Registry
- 6
Phase 5 — MCP Integration
Open Source
Anyone can build on MAK4I
The specification is public and MIT licensed, with a working reference implementation. Anyone can implement it; Talvik builds commercial infrastructure and services on top, not a wall around it.
Public specification
The MAK4I spec is developed in the open, and anyone can build a conforming implementation.
MIT licensed
Free to use, adapt, and build on without restrictive licensing terms.
Community driven
Direction is shaped by feedback from the developers who use it.
Get Started
Try the MAK4I Developer Preview
The protocol and specification are public on GitHub. The hosted reference implementation is access-controlled — email us and we'll set you up with credentials.