- Home
- Skills
- Productivity
- opencode brain
Works with the AI Agents you already use
opencode brain
A persistent, file-based memory and self-distilling context architecture for AI coding agents.
$20
opencode brain
Example session with this skill installed
Set up OpenCode Brain on my machine from the template, interview me for my preferences and projects, and wire it into my agent configuration.
- Read your context and instructions
- Compiled the opencode brain
Summary
Cloned opencode-brain, initialized personal memory files via tools/init.mjs, personalized your core profile, and registered AGENTS.md into your OpenCode instructions configuration.
Installation & Initialization
| Step | Action | Status | Output / Path |
|---|---|---|---|
| 0. Platform | OS & Runtime Check | PASS | Node v22.x, cross-platform paths |
| 1. Init | node tools/init.mjs | PASS | Created local gitignored memory files from templates |
| 2. Interview | Captured identity & rules | PASS | Populated PROFILE.md, CHANNELS.md, SYSTEM.md |
| 3. Config Wire | OpenCode opencode.json | PASS | Appended AGENTS.md to instructions array |
| 4. Verification | Read-order diagnostic | PASS | Agent verified read sequence and surface formatting |
Active Memory Structure
PROFILE.md– Your communication style (concise, professional, no fluff) and commit rules.CHANNELS.md– Tailored responses for Telegram (chat style) vs CLI/TUI (terminal style).PROJECTS.md– Workspace directory mappings.DECISIONS.md– Append-only log for framework and architectural choices.journal/2026-10.md– Timestamped log for continuous learnings.
Daily Usage Commands
# Record a preference or tech decision
node tools/remember.mjs "User prefers strict TypeScript with NodeNext resolution" --tag preference
# Review journal and promote repeated truths into permanent docs
node tools/learn.mjs review
# Scaffold a new procedural skill
node tools/learn.mjs new deployment
Connects securely to your tools. The creator never sees your data.
About this skill
Your AI coding agent forgets everything between sessions. AGENTS.md fixes that for one repository—OpenCode Brain fixes it for your entire machine.
It provides a single directory of structured Markdown stored outside your projects that survives session restarts, IDE reloads, and new repositories. It informs your agent who you are, what you are building, what technical decisions you have already made, and what it must never touch.
Zero database, zero daemon, zero runtime dependencies. Pure Markdown powered by lightweight Node.js helper scripts.
Multi-Layered Architecture: Progressive Disclosure
Most memory systems bloat context windows with thousands of irrelevant tokens. OpenCode Brain uses strict token bounding:
- Always-on Layer (
AGENTS.md): Injected into every session via agent config. Deliberately kept short (<100 lines) to keep per-prompt token costs minimal. - On-Demand Context:
PROFILE.md: User identity, communication rules, constraints.CHANNELS.md: Per-surface rules: Telegram chat style vs CLI/TUI terminal style.SYSTEM.md: Machine topology, paths, toolchains, active daemons.PROJECTS.md: Active project registry, mappings, and stack guidelines.DECISIONS.md: Immutable append-only ledger of architectural and tech choices.SKILLS.md&skills/<class>/: Modular procedures loaded only when a matching task runs.journal/YYYY-MM.md: Dated history of daily learnings and corrections.
The Self-Improving Growth Loop
- Capture: Store preferences, corrections, and root-causes as single journal entries (
node tools/remember.mjsor Telegram/remember). - Review: Review journal trends and promote repeated truths into durable files (
node tools/learn.mjs review). - Distill: Procedures repeated 3 times become modular reusable skills (
node tools/learn.mjs new <class>). - Retire: Stale instructions are retired when tools update.
Requires local file system access. Memory management is file-based only, not database-driven. User must manually register the root memory path in their agent configuration.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 25 days ago
Requires
Permissions
File Scopes
Purely local filesystem operations. No network ports, no external API calls, and no background daemons.