Skip to content

Personal AI context

SelfContext

Think with context you own.

People repeatedly explain the same history, goals, projects, preferences, decisions, and constraints to AI systems. SelfContext keeps the context worth carrying forward in portable Markdown, then lets the AI harness you already use work from it.

View on GitHubOpen source · MIT

Why I built it

I built SelfContext because I was tired of rebuilding context every time I wanted to use AI for something meaningful. After using it regularly, I realized the important part wasn't memory by itself. It was being able to continue thinking from context I had deliberately kept.

I use it as a brainstorming companion when I want to carry a line of thought across sessions. The point is not that it knows me; it gives the AI tool a deliberate context to work from.

How it works

SelfContext does not ship another AI interface or runtime. Natural-language interaction from the repository root remains the primary UX.

User
  ↓
Existing AI harness + model
  ↓
SelfContext skills
  ↓
Local Context Vault
(portable Markdown)

Start here

Install it where you work.

Open the repository root in an AI harness that supports project-local Agent Skills. SelfContext initializes the local Context Vault on first use, and an existing vault can be copied there. It is not another AI interface or runtime; no database, embeddings system, or slash command is required.

git clone https://github.com/joacod/self-context.git
cd self-context

Use it

Continue from the repository root

Use natural language from the repository root. The existing AI harness and model provide execution and intelligence; SelfContext skills handle ingestion, retrieval, review, reasoning, and lifecycle maintenance for the context you choose to keep:

ingest my resume into SelfContext
what does my context say about [topic]?
help me think through [decision] using my context
review my context for stale or conflicting information
based on my career context, how should I position myself for [role]?

Why this is different from a memory dump

SelfContext maintains context instead of blindly accumulating memories. User-stated or confirmed information, source-derived information, agent observations and inferences, and derived conclusions remain distinguishable. Stale context, contradictions, and unresolved questions stay visible for review instead of becoming one undifferentiated pile.

What stays yours

  • The local Markdown vault is canonical and remains readable and editable without SelfContext.
  • The context is portable across models and compatible AI harnesses.
  • There is no hosted SelfContext memory service, database, embeddings system, custom runtime, or SelfContext chatbot.
  • The model or provider can still see information you supply to that tool.
  • You remain the final authority over durable context, including what to inspect, correct, copy, or delete.

Current scope

Current scope includes focused areas such as career, learning, writing, relationships, media/taste, and ventures/projects. They are not separate products; they share the same context lifecycle.