Axiomux

The description states that Axiomux is:

• An MIT-licensed control plane for coding agents
• A system that unifies provider login and model selection while maintaining project continuity, steering, governed memory, quarantined skills, effect approvals, fallback, and tool-less multi-model Fusion under one operator-controlled runtime
• Not an OMP fork or extension but rather owns the project registry, sessions, steering, follow-up queue, local tools, effect journal, memory, reusable skills, workflows, fallback, Fusion/MoA, and bilingual local UI
• Implemented using TypeScript, Node.js 24, SQLite/FTS5, and a bilingual local HTML/CSS/JavaScript UI

Inference: Axiomux appears to be a local runtime environment that manages agent interactions with multiple AI providers while preserving state and enforcing safety controls. It is not a hosted service but rather a client-side tool.

The description states:

• Axiomux positions itself as enabling switching between AI providers without surrendering project memory, tool governance, or operator control
• It separates the model-execution substrate from the agent's continuity and authority
• The author claims to have made core product decisions such as creating a new harness rather than a mod, keeping strict isolation and explicit approval, preserving fallback without replaying effects, making Fusion arbitrary but tool-less, and separating general engineering improvements from submission-only work

Inference: The positioning evolved from addressing fragmentation in coding agent ecosystems to proposing a unified control plane that maintains operator sovereignty over project state and execution flow.

The description states:

• Axiomux is designed for users who want to switch AI providers without losing context or control
• It targets developers working with coding agents across multiple models
• The system runs as a pinned, digest-verified external process for supported authentication and model execution

Not evidenced: No specific customer segments, personas, or use cases beyond general developer needs are identified.

The description states:

• Axiomux is MIT licensed
• It includes a judge package with deterministic offline demo that executes real production components without credentials, network access, or user-state mutation
• The system is packaged as an npm tarball with explicit licensing, provenance, threat model, and known issues

Not evidenced: No pricing information, monetization strategy, or revenue model is provided.

The description states:

• Built with TypeScript, Node.js 24, SQLite/FTS5, and a bilingual local HTML/CSS/JavaScript UI
• Implements deterministic offline demo that executes real production components without provider credentials, network access, or user-state mutation
• Memory is stored locally with provenance, review state, feedback, relations, and bounded retrieval
• Automatic learning creates candidates requiring human review, converts them into quarantine, requires second approval, and records evaluation
• Consequential tool effects are journaled before execution
• Cross-provider fallback stops once an effect makes replay unsafe
• Fusion consultants are tool-less and failures are isolated; only the primary agent can request actions

Inference: The technical architecture suggests a strong emphasis on local execution, safety through human review, and deterministic behavior. However, these claims remain unproven in practice.

The description states:

• Axiomux is an "honest alpha"
• Not a claim of universal provider support or a production sandbox
• Catalogue entries, account entitlement, and successful model turns are reported separately
• Browser/desktop adapters remain experimental
• The project was submitted to the OpenAI 2026 hackathon

Not evidenced: No evidence of revenue, customers, user adoption, or product-market fit beyond the author's own description.

The description states:

• Coding agents increasingly offer many models, but switching providers often means switching entire operational context
• Axiomux separates model-execution substrate from agent continuity and authority

Not evidenced: No mention of competitors, market positioning relative to existing tools, or competitive advantages beyond the stated claims.

The description states:

• Reusing OMP's login surface without becoming an OMP fork or inheriting extension authority
• Keeping fallback useful while preventing retries from repeating external effects
• Making memory and self-learning valuable without treating imported text or generated code as trusted
• Packaging the compiled CLI and installer so a fresh consumer installation needs no TypeScript toolchain
• Distinguishing catalogue capability, successful authentication, model availability, and a real turn in every public claim

Red flags:

• The system is described as an "honest alpha" with no production-ready claims
• Claims about safety mechanisms (e.g., two-stage human review) are unproven in practice
• No evidence of actual user testing or feedback loops
• The project is a single-person effort, raising questions about scalability and long-term maintenance

1. How does Axiomux actually implement effect journaling and ensure that replay-safe fallbacks are enforced?
2. What are the specific mechanisms for quarantining learned skills and ensuring human review occurs before code activation?
3. Can you demonstrate a working example of switching between two different AI providers while preserving project state?
4. How does Axiomux handle cross-provider compatibility issues, especially around model APIs and token formats?
5. What is the current status of browser/desktop automation support, and how does it differ from other platforms?
6. Are there any known limitations or trade-offs in terms of performance, latency, or feature completeness?