OneContext - Agent Self-Managed Context Layer

A persistent context layer that enables AI coding agents to retain and reuse project knowledge across sessions, devices, and collaborators.

What Is It?

OneContext is a self-managed contextual layer designed to overcome the most common issue in AI-assisted programming: context loss. It's a persistent environment that AI coders can easily reuse, allowing work to continue without having to re-explain requirements, decisions, or previous progress.

The Problem: AI coding agents suffer from session-based memory. Each new session often starts with partial or no memory of prior decisions, architecture, or progress. Developers must constantly restate requirements and re-establish project understanding.

The Solution: OneContext provides a durable, shareable context layer that makes project memory persistent across:

Sessions
Devices
Coding agents (Codex, Claude Code, etc.)
Team members

What Problem Does It Solve?

Context Fragmentation Issues

Repeated explanations of project structure and goals
Inconsistent outputs from different sessions
Lost architectural decisions and design choices
Difficult collaboration - can't share exact AI configuration
Knowledge decay in long-running projects

Why Persistent Context Matters

Modern AI coders are efficient but depend on session-based memory. After a session ends, its context is lost until manually restored. This creates friction and delays in development.

How It Works

1. Persistent Context Across Sessions

When you start an AI coding agent using OneContext, it collects relevant project details and history into a unified context layer. This layer is reused each time a new session launches within that project.

No need to:

Copy/paste summaries manually
Rephrase criteria
Re-explain project structure

The agent already has access to the entire project understanding.

2. Multiple Agents, One Shared Memory

OneContext lets you run multiple AI agents in the same context. Each agent has access to the entire project memory, enabling concurrent or sequential workflows without compromising continuity.

Use cases:

Switching between different AI coding agents
Restarting tools during long development cycles
Running parallel agents on different features

Share the entire project context via a hyperlink. Anyone with the link can work with the same information and see exactly what the agent knows about the project.

This transforms context from:

❌ Private session-bound artifact
✅ Shareable, collaborative asset

Quick Start

# Install globally
npm i -g onecontext-ai

# Run
onecontext

# Or use short alias
oc

Commands:

# Check version
onecontext version

# Show help
onecontext --help

# Update
onecontext update

# Troubleshoot
onecontext doctor --fix-upgrade

Key Features

Feature	Description
Persistent Context Layer	Maintains project memory across sessions and devices
Multi-Agent Support	Multiple AI agents can operate under the same context
Automatic Context Management	No manual summaries or re-prompting required
Shareable Context Links	Enables collaboration using identical project memory
CLI-Based Workflow	Fits naturally into developer tooling
Agent Trajectory Recording	Records your agents' paths and decisions

Supported Agents

Works with major AI coding agents:

Claude Code
Codex
Other AI coding agents

Practical Use Cases

1. Long-Running Software Projects

For projects lasting weeks or months, OneContext prevents knowledge decay between sessions. Pause and resume work without reconstructing context.

2. Switching Between Devices

Moving between machines? The persistent context ensures the AI agent is fully aware of the project's current state.

3. Team Collaboration

Share context links to enable team members to take on work using the same knowledge. Reduces onboarding time and confusion.

4. Experimentation and Prototyping

Explore different solutions with distinct agents without losing the project's core context.

5. Slack Integration

Anyone can talk to your agents on Slack using the shared context.

Benefits

Reduced Repetition - Eliminates need to explain project details repeatedly
Better Consistency - Agents generate more consistent outputs over time
Saves Time - Faster onboarding for new sessions or collaborations
Scalable Workflows - Supports complex, multi-agent development patterns
True Continuity - Agents behave like long-term partners, not one-shot tools

OneContext vs Traditional Workflows

Aspect	Traditional Workflow	With OneContext
Session Memory	Temporary	Persistent
Context Reuse	Manual	Automatic
Collaboration	Re-explain or document	Share context directly
Multi-Agent Continuity	Fragmented	Unified

Limitations and Considerations

While OneContext solves persistence, some practical considerations remain:

Context quality depends on project history quality
Large-scale projects might require disciplined use to reduce noise
Access control matters when sharing context links
Storage/performance limits may vary based on usage

Prerequisites

Node.js >= 16
Python 3.11+ with one of: uv, pipx, pip3, or pip

Who Built It

Created by Junde Wu (@JundeMorsenWu)

Junde is interested in Buddhism, AI, Medicine, Economics, and Modern Art. His academic work has been published in CVPR, ECCV, IEEE TIP, AAAI, MICCAI, MedIA.

Personal note from the creator:

"I built it for myself but now I can't work without it, so it felt wrong not to share."

Why It Matters

As AI development workflows become more complex and layered, persistent contexts like OneContext are expected to become foundational infrastructure. Tools that preserve project memory, minimize repetition, and enable seamless collaboration will play a greater role in how developers build and maintain software.

The shift: OneContext moves AI coding agents from "one-shot tools" to "long-term partners."

Philosophy

"Instead of treating every interaction as an isolated exchange, it enables agents to retain and reuse project knowledge across sessions, devices, and even collaborators."

This aligns with broader trends in AI agent development:

steve-yegge's Gas Town - agent factories need coordination
temporal-io-ai - workflows need persistence
Spec-driven development (github-spec-kit, openspec.dev) - context needs structure

OneContext fills the gap between sessions, making context a first-class citizen in AI-assisted development.

steve-yegge - Gas Town uses Beads for agent coordination (similar persistence problem)
temporal-io-ai - Workflow orchestration with durable state
github-spec-kit - Spec-driven development with persistent artifacts