GitHub - grigoriitropin/vehir-platform: An AI-native, self-hosting platform for autonomous agents, with a user-space microkernel, worker runtime, native compiler, and content-addressed generations. · GitHub
/" data-turbo-transient="true" />
Skip to content
Search or jump to...
Search code, repositories, users, issues, pull requests...
-->
Search
Clear
Search syntax tips
Provide feedback
--><br>We read every piece of feedback, and take your input very seriously.
Include my email address so I can be contacted
Cancel
Submit feedback
Saved searches
Use saved searches to filter your results more quickly
-->
Name
Query
To see all available qualifiers, see our documentation.
Cancel
Create saved search
Sign in
/;ref_cta:Sign up;ref_loc:header logged out"}"<br>Sign up
Appearance settings
Resetting focus
You signed in with another tab or window. Reload to refresh your session.<br>You signed out in another tab or window. Reload to refresh your session.<br>You switched accounts on another tab or window. Reload to refresh your session.
Dismiss alert
{{ message }}
grigoriitropin
vehir-platform
Public
Notifications<br>You must be signed in to change notification settings
Fork
Star
main
BranchesTags
Go to file
CodeOpen more actions menu
Folders and files<br>NameNameLast commit message<br>Last commit date<br>Latest commit
History<br>11 Commits<br>11 Commits
ARCHITECTURE.md
ARCHITECTURE.md
INSTALL.md
INSTALL.md
LICENSE
LICENSE
README.md
README.md
View all files
Repository files navigation
Vehir Platform
Vehir Platform is an experimental AI-native computing environment designed around agent–computer interaction. In the same way that conventional systems use human–computer interaction principles to make computing accessible to people, Vehir designs the entire software environment around the needs of AI agents.
Its interfaces and tooling are designed around structured, discoverable input and output that AI agents can use directly. Vehir explores how the software environment itself can be made more convenient for machine reasoning, rather than treating agent support only as an additional interface over existing workflows.
The current implementation provides the foundations of this environment, including structured agent tools, IPM packages, native compilation and loading, content-addressed generations, reconciliation, and supervised services. The platform is still under active development.
System requirements
Linux
x86-64 processor
4 GB of RAM or more recommended
Installation
Copy the following and send it to your AI agent:
Read the complete installation guide at<br>https://github.com/grigoriitropin/vehir-platform/blob/main/INSTALL.md<br>before taking any action. Follow it from beginning to end, and ask the<br>user whenever it requires a choice or information about the host system.
Why AI-native
Vehir is built for machine-to-machine interaction. Instead of wrapping<br>existing human-facing tools with AI interfaces, it designs every layer<br>to be directly usable by agents:
Structured input and output reduce reliance on free-form text parsing.
Self-discoverable tools — the system tells the agent what it can do.
Machine-readable errors with remediation hints.
The source language (IPM) is designed for machine writing and verification, not manual editing.
Declarative build — the agent declares desired state; the system reconciles.
Some features are still under development and may not yet be complete.
Core systems
Vehir is more than an agent tool server. It combines several layers into<br>one self-hosting platform:
User-space microkernel and worker runtime. The kernel launches and<br>manages workers using arenas, shared memory, and doorbell-based<br>coordination. Managed services use the same worker model.
Self-hosting native compiler. IPM packages are compiled directly to<br>x86-64 native code. The compiler builds itself, with fixpoint verification<br>across generations.
Native loading. Vehir resolves object closures, applies relocations,<br>and loads its own compiled artifacts without relying on a conventional<br>application runtime.
Content-addressed storage. Sources and compiled objects are addressed<br>by digest, deduplicated, and recorded in immutable generations.
Declarative reconciliation. A declared build state and package graph<br>are transformed into a verified generation, followed by an atomic<br>current switch.
Code-derived agent tools. MCP operations, request contracts, help, and<br>request validation are derived from the tool implementations rather than<br>maintained as separate handwritten interfaces.
Whole-universe enforcement. Cross-package analysis checks structural<br>and policy invariants before a generation is promoted.
Source availability
This GitHub repository contains documentation only. Release archives are<br>not binary-only distributions: every published generation archive<br>(N.tar.gz) contains the complete IPM source tree together with the native<br>artifacts needed to bootstrap that exact generation. The source...