CliDeck MCP — Network Knowledge

The description states that CliDeck MCP is an open-source, agent-native framework for verified, continuously updated MCP knowledge systems. It uses a Model Context Protocol (MCP) architecture to separate exact, rapidly changing knowledge from a stable LLM core.

It includes:

• A core system managing immutable revisions, rollback capability, provenance, and conflict detection.
• Domain Packs, which define schemas, normalization rules, and mappings into universal knowledge revisions.
• A Network Knowledge Domain Pack currently covering Cisco Catalyst and IOS-XE devices.
• A continuous learning pipeline powered by Codex that handles discovery, conversion, OCR, chunking, Luna analysis, verification, and publication.

The system is built with:

• Node.js 24, TypeScript, Hono, Zod 4, MCP TypeScript SDK
• PostgreSQL 16
• Cloudflare Tunnel, systemd services
• GitHub Actions, Codex integration

Inference The framework appears to be designed for technical domains like networking, where precision and version control are critical. It is not a general-purpose AI assistant but rather a knowledge infrastructure component.

The author states:

• General-purpose AI models are unreliable on exact commands, product versions, limitations, tolerances, and operational procedures.
• The system keeps the fundamentals of an LLM (reasoning ability, tool-use skills) while placing exact knowledge in a verified external system.
• This architecture avoids retraining models every time documentation changes.

Inference The positioning evolved from solving the problem of unreliable AI responses in technical domains to building an infrastructure for reusable, verified knowledge systems that can be extended via Domain Packs.

The description states:

• The system supports networking engineers and operations teams who need precise CLI commands, diagnostics, workflows, upgrade guidance, and risk management.
• It is designed for technical domains with high precision requirements, such as network device configurations.
• The public demo exposes 16 MCP tools across Network and generic Domain Pack interfaces.

Inference The primary ICP appears to be technical professionals working in enterprise networking environments, particularly those using Cisco, Junos, or Arista devices. It targets users who need reliable, up-to-date technical knowledge without relying on potentially inaccurate LLMs.

The description states:

• The project is open-source under Apache-2.0 license.
• Source code and fixtures are distributed publicly.
• Production knowledge, third-party documents, private manuals, and operator-imported datasets are not included in the repository.
• No pricing or monetization model is mentioned.

Inference There is no evidence of a commercial business model or pricing structure. The project appears to be open-source with no stated revenue path.

The description states:

• Built using Node.js 24, TypeScript, Hono, Zod 4, MCP TypeScript SDK.
• PostgreSQL 16 as the only required stateful dependency.
• Uses systemd services behind Cloudflare Tunnel.
• Implements continuous isolated Codex/Luna executors with PostgreSQL leases.
• Supports atomic rollback, immutable delta releases, checkpoints.
• Includes a complete operations console (not just backend endpoints).
• Public demo uses the same compiled React app as admin interface.

Inference The technical stack is modern and well-suited for scalable, secure delivery. The system supports both public read-only access and private administrative functions. It shows strong engineering maturity in terms of observability, security, and modularity.

The description states:

• Production instance contains more than 66,000 active knowledge revisions.
• Public product eval includes 250 scenarios and passes all with zero dangerous false-safe outcomes.
• More than 99% of completed answers use the deterministic no-AI path.
• The system supports a public MCP endpoint, live demo, source code repo, and product page.

Inference There is evidence of a functional, production-ready system with significant internal knowledge coverage. However, there is no evidence of customer adoption, revenue, or user base beyond the author’s own claims.

The description states:

• Most retrieval systems either place an LLM in every response path or require custom ingestion apps per subject.
• CliDeck MCP separates a stable knowledge core from subject-specific Domain Packs.
• It is not a "network database disguised as a universal framework."

Inference The competitive landscape includes traditional RAG systems, proprietary knowledge bases, and custom-built ingestion pipelines. The key differentiator is the separation of exact knowledge from LLM reasoning and the use of deterministic validation.

• No evidence of traction or revenue: The system is described as self-built with no mention of customers, users, or monetization.
• Single-person team: Only one member listed (Valentyn M), which may limit scalability or long-term maintenance.
• Dependency on Codex: While the pipeline can run locally via Codex, it relies on the operator's plan and usage limits — not a scalable model for third-party access.
• Limited domain coverage: Currently focused on Cisco devices; other vendors have limited coverage.
• Self-reported metrics: All performance data is self-reported without independent verification.

1. What is the actual source of the knowledge content (e.g., vendor manuals, internal documentation)?
2. How do you plan to scale beyond a single developer's capacity?
3. Are there any partnerships or integrations with vendors like Cisco, Junos, or Arista?
4. Has anyone else used this system in production outside of your own use case?
5. What is the long-term vision for monetization or commercial adoption?
6. How do you ensure quality and consistency across different Domain Packs?
7. Can you provide evidence of how the system handles edge cases or ambiguous inputs?