Introducing MAK4I
Today, Talvik is releasing the first public preview of MAK4I — an open protocol for reusable AI artifacts and interoperability across AI systems.
MAK4I v0.1.0 is now public.
This is Phase 0: the protocol, specification, standards, architecture, governance model, and roadmap are open for discussion and community feedback.
The Problem
AI is increasingly becoming part of the entire software lifecycle.
Developers use AI coding tools.
GitHub has AI capabilities around development workflows.
Issue and project management platforms are adding AI.
Documentation platforms have AI assistants.
Operations and infrastructure platforms are becoming AI-assisted.
But these systems do not necessarily share a common model for the reusable work created through them.
Useful context, decisions, standards, procedures, skills, project knowledge, and other artifacts can remain fragmented across tools.
That creates a simple question:
If useful work has already been created with AI, why should another AI system have to reconstruct it?
What MAK4I Is
MAK4I proposes a protocol-level model for reusable AI artifacts.
Rather than treating AI context as something that exists only inside a conversation or a vendor-specific memory system, MAK4I defines artifacts that can be represented, versioned, discovered, governed, and reused.
Examples can include:
- project context
- organizational standards
- procedures and workflows
- reusable prompts
- skills
- code patterns
- architecture decisions
- operational runbooks
- other reusable AI context
The goal is not simply to store more memory.
The goal is to make useful AI work reusable and interoperable.
Reuse Before Regeneration
One of the core ideas behind MAK4I is straightforward:
Before generating something again, determine whether an appropriate reusable artifact already exists.
This changes the model from:
Generate → use → forget → regenerate
to:
Discover → reuse → evolve
When something useful already exists, an AI system should be able to discover and reuse it rather than reconstructing the same context or work repeatedly.
MAK4I and MCP
MAK4I is not intended to compete with the Model Context Protocol (MCP).
They operate at different layers.
MCP provides a standard way for AI applications to connect to external tools and data.
MAK4I focuses on how reusable AI artifacts are represented, governed, discovered, reused, and evolved.
A future MAK4I MCP server is planned so MCP-compatible tools can access MAK4I artifacts through a standard integration path.
MCP is therefore one integration mechanism for MAK4I, rather than a dependency on which the protocol itself relies.
Open Protocol
The MAK4I protocol is released under the MIT License.
Talvik intends to keep the protocol open while building commercial products and services around the ecosystem over time.
The idea is similar to other open-technology ecosystems: adoption of the open layer creates opportunities for tooling, hosted infrastructure, enterprise capabilities, support, and services around it.
Current Status
MAK4I is currently Phase 0 / Public Preview.
The v0.1.0 release establishes the initial public foundation, including:
- protocol specification
- artifact standards
- architecture
- governance
- conformance model
- security policy
- competitive landscape
- roadmap
- planned developer experience
The CLI, SDKs, hosted registry, reference API, and MCP server described in the roadmap are planned capabilities and are not production interfaces today.
That distinction is intentional.
We are publishing the protocol early so its assumptions can be challenged before the implementation becomes difficult to change.
What's Next
Phase 1 focuses on developing the protocol specification further, including areas such as:
- artifact metadata
- context injection
- memory and artifact resolution
- conflict resolution
- versioning and compatibility
- community RFCs
Later phases introduce the reference implementation, APIs, CLI, SDKs, hosted registry, and MCP integration.
Why Publish This Now?
Because interoperability protocols benefit from being discussed before one implementation defines everything.
MAK4I should not be designed solely around Talvik's assumptions.
We want developers, AI engineers, platform engineers, architects, and others working with AI systems to challenge the model.
What is missing?
What doesn't belong at the protocol layer?
Where would interoperability break?
What should an AI system be allowed to reuse automatically?
How should conflicting artifacts be resolved?
Those questions are more valuable right now than pretending the protocol is finished.
Get Involved
The MAK4I repository is public and the first release is available now.
Website:
https://talvik.ai
GitHub:
https://github.com/talvikai/mak4i-protocol
GitHub Discussions can be used for broader protocol ideas and questions.
Issues can be used for specific problems and proposed improvements.
Pull requests are welcome for concrete contributions.
Security vulnerabilities should be reported privately according to SECURITY.md rather than through public issues.
MAK4I
The protocol for not rebuilding what your AI already built.
— Talvik, Inc.
Top comments (0)