Knowledge Management
made ridiculously simple for humans and AI.
Turn scattered documentation into living, structured knowledge that your AI agent understands.
1. Feed the knowledge base with your files and thoughts
Chat with the AI agent to capture your ideas and thoughts into the knowledge base, or feed it unstructured, narrative documents and files.
2. AI models & validates
cogNNitive analyzes your files and thoughts to identify entities, concepts, and elements, structuring and storing them as modelsβeither from domain templates or built from scratch.
3. Query & generate artifacts
Query your knowledge base by chatting with an AI agent, or ask it to generate deliverables like presentations, web pages, and documents.
What is cogNNitive?
An open, unified ecosystem designed to make documentation readable and editable by humans and AI.
iNNfo β Knowledge Modeling
A simple specification chain for structuring Markdown documents. Clear, validated, and engine-backed.
actioNN β AI Agent Skills
Modular capabilities that teach your OpenCode AI agent specialized domain workflows automatically.
iNNfo Modeler β Visual App
A browser-based workspace editor to view, edit, and validate knowledge graphs without installing servers.
Ecosystem & Information Flow
Complete vertical integration: from ingestion to engine validation, web access, and artifact generation.
flowchart TD
subgraph Step1["Step 1: Ingestion & Setup"]
U["π€ User in OpenCode"] -->|Bootstrap Prompt| Router["β‘ actioNN Skills (Router & traNNsform)"]
RawDocs["π Raw Documents (PDF, DOCX)"] --> Router
end
subgraph Step2["Step 2: Validation & Engine"]
Router --> Models["π Structured Models (_NN.md)"]
Models <--> MCP["βοΈ innfo-mcp Server + Core Engine"]
end
subgraph Step3["Step 3: Delivery & Access"]
MCP <--> Modeler["π οΈ iNNfo Modeler App (Web Editor)"]
MCP <--> AgentChat["π¬ OpenCode Conversational Queries"]
end
subgraph Step4["Step 4: Artifact Generation"]
Modeler & AgentChat --> Artifacts["π Final Deliverables\n(Visual Dashboards, Exec Summaries, Scripts)"]
end
classDef s1 fill:#FAFAFC,stroke:#636366,color:#111112;
classDef s2 fill:#F6EEF6,stroke:#4D0E4E,color:#4D0E4E;
classDef s3 fill:#FAFAFC,stroke:#6A1B6B,color:#111112;
classDef s4 fill:#4D0E4E,stroke:#4D0E4E,color:#FFFFFF;
class U,Router,RawDocs s1;
class Models,MCP s2;
class Modeler,AgentChat s3;
class Artifacts s4;
6 Key Benefits
Why modern teams and AI developers choose cogNNitive for their documentation.
Zero Vendor Lock-in
Plain text Markdown files stored in your own Git repository. You own your knowledge forever.
AI That Never Guesses
Deterministic validation guarantees your OpenCode agent always works with reliable data.
No Setup Friction
Works directly inside OpenCode Desktop with a single prompt. No complex setup required.
Always Up to Date
Detects structural drift and outdated information automatically before it causes mistakes.
Visual & Flexible
Edit visually in your browser app or textually through conversational AI instructions.
100% Free & Open Source
Built for the open community under the MIT license. No hidden subscriptions or API costs.
How to Use It
Get started with cogNNitive in 3 simple steps inside OpenCode Desktop.
Open OpenCode Desktop
Launch OpenCode Desktop on your computer and open your project workspace folder.
Prompt Your Agent
Tell your AI agent the single bootstrap phrase in the chat:
Enjoy Living Knowledge
OpenCode automatically installs the skills, configures tools, and presents your interactive workflow menu.
The A ⇄ K Paradigm
Turn Anything into structured Knowledge, and back into Anything using AI.
flowchart LR
A1["📄 Anything (A)
White light β your original file"] -->|"A →"| Prism["📡 AI Prism
The intelligence loop"]
Prism --> K["🌈 Structured Knowledge (K)
The spectrum β entities, relations, metadata"]
K -->|"K →"| A2["🎨 Anything (A)
A new ray β the file you choose"]
A2 -.->|"A ⇄ K
virtuous cycle"| A1
classDef a1 fill:#FAFAFC,stroke:#636366,color:#111112;
classDef k fill:#F6EEF6,stroke:#4D0E4E,color:#4D0E4E;
classDef a2 fill:#4D0E4E,stroke:#4D0E4E,color:#FFFFFF;
class A1 a1;
class Prism,K k;
class A2 a2;
Anything In (A)
Any file you have: PDF, DOCX, meeting notes, spreadsheet, or raw text. This is your white light.
The AI Prism (⇄)
Not a shallow format converter. AI decomposes the original file into a structured knowledge layer β entities, relations, and metadata.
Anything Out (A)
AI re-composes that knowledge into the new file you need β dashboard, summary, spec, or script. The color you choose.
The cycle is fully reversible (A ⇄ K): inputs and outputs stay decoupled through a single semantic core, and every pass through the prism enriches the knowledge behind the file.
Goodbye Fragmented Documentation
The difference between stale files nobody trusts and an active semantic foundation.
# Architecture Doc (Updated: 2023?)
TODO: Update this section after rewrite
See link: http://broken-notion-link.internal
Note: Might not match current database schema.
Ask Dave if this still applies.
---
provenance: { hash: "sha256:4a8f9...", verified: true }
spec_version: "V_0-1-0"
mcp_tools: ["innfo_search", "innfo_trace_matrix"]
---
| ID | Requirement | Matrix Relation | Verified |
| M-1 | Living Schema | ACT-004 | [x] True |
What cogNNitive is NOT
Clear boundaries keep the ecosystem honest, simple, and yours. If it isn't listed here, it isn't the product.
Not a database
Models are plain Markdown files in your own Git repository. No proprietary storage engine, no hidden silo, no lock-in.
Not a hosted platform
No mandatory cloud service or managed infrastructure that owns your knowledge. Everything runs locally or in your browser.
Models never execute code
_NN.md files are data, not programs β no macros, scripts, arbitrary commands, or auto-installed plugins. Opening a model never runs anything.
Not a real-time collaboration platform
No live multi-user editing, presence, or sync protocol. Collaboration is the engineering way: files in Git, branches, and reviews.
Not a vector database / RAG platform
It structures knowledge so retrieval works, but it does not store embeddings or manage retrieval infrastructure. Connect the vector tool of your choice.
Not a document authoring suite
It does not replace your wiki, CMS, or word processor. It organizes the knowledge those tools produce into validated, linked models.
Not a universal format
OKF-compatible and plain Markdown, but not the single format for all knowledge. Your source of truth stays yours.
Not a one-shot AI converter
The A ⇄ K cycle is reversible and iterative. A single unvalidated lossy conversion is not the product β living, validated models are.