FASTEST PATH TO VALUE

Get Started in 3 Minutes

You do not need to read through manuals or run complex configuration scripts to get started. Here is the exact 3-step setup to lock in your AI memory and portable skills today:

1
Create your vault and copy your link

Sign up at /auth/signup and copy your permanent Vault Address: https://app.skillvault.dev/v/<your-username>.

2
Connect your tool in one step

Choose how you interact with AI:

  • Desktop Coding Agents (Claude Code, Cursor, Windsurf, Antigravity): In your agent chat, type: Set up my skillvault https://app.skillvault.dev/v/<your-username> and click Authorize in your browser.
  • Web & Mobile Chats (ChatGPT, Claude.ai, Kimi): Add your vault as an MCP connector, or paste your private read URL into the conversation.
3
Prompt your AI and never explain twice

Your agent now checks your vault on Step 0. When you establish a rule, fix a bug, or commit code, your agent logs it automatically. Switch between your laptop, terminal, and phone with zero amnesia.

01 On a Laptop or PC You Own (CLI Setup)

When working on your desktop or laptop, SkillVault connects directly to your AI agent tools through its local CLI utility. Setup only needs to be completed once per machine. After pairing, any AI agent on your computer can access your vault and execute skills seamlessly.

One-Time Setup Instructions

  1. Go to your SkillVault Dashboard and copy your Vault Address (https://app.skillvault.dev/v/your-username).
  2. Open your favorite AI coding tool or CLI agent (e.g. Claude Code, OpenCode, Cursor, Antigravity) and send a simple message:
    "Set up my SkillVault at https://app.skillvault.dev/v/your-username"
  3. Your primary web browser will automatically open to app.skillvault.dev's authorization page. (If you are not already logged into skillvault.dev in your browser, log in first), then click the green Authorize This Device button.
  4. That's it! Your machine is paired. Every AI agent on your laptop can now discover, read, and run skills from your vault.

Supported AI Agent Harnesses

SkillVault automatically integrates with all major AI coding tools and desktop agent harnesses out of the box:

• Claude Code
• Cursor
• Windsurf
• Antigravity
• OpenCode
• Codex / CommandCode
• OpenClaw
• Aider
• Cline
• Kilo Code
• Hermes
• DeepSeek Harness
📬 Missing Your Favorite AI Agent?

If you use an AI coding tool or agent framework not listed above, email us at support@skillvault.dev and we will add native support to the CLI!

02 On an AI Chatbot Account You Log Into (Connectors & MCP)

You can connect SkillVault directly to your AI chatbot accounts (such as ChatGPT, Claude.ai, or Perplexity). Once connected to your account, SkillVault is accessible from anywhere you log into that account, whether on a web browser, a tablet, or a secondary work laptop.

Connecting Your Accounts

A. ChatGPT:

  • In ChatGPT, locate Plugins in the left-hand menu.
  • Click the + (Plus sign) to add a New Plugin.
  • Select Server URL and enter https://app.skillvault.dev/mcp.

B. Claude.ai:

  • In Claude.ai, click the + (Plus sign) in the chat bar.
  • Select Add Connector → Add Custom Connector.
  • Paste https://app.skillvault.dev/mcp (or your MCP / Connector URL from your Dashboard) and save.

C. Perplexity:

  • In Perplexity, click the + (Plus sign) in the chat bar.
  • Select Connectors → Manage Connectors → Add Custom Connector.
  • Set Host to app.skillvault.dev and Authentication Type to Bearer token.
  • Paste your Permanent Token from /dashboard/token into the Token field.
  • In the Notes (Instructions) box, paste:
    PERPLEXITY COMPUTER INSTRUCTIONS
    List skills: GET https://app.skillvault.dev/api/v1/me/skills Get skill: GET https://app.skillvault.dev/api/v1/me/skills/{name} Run proxy: POST https://app.skillvault.dev/api/v1/me/skills/{name}/proxy

03 On a Mobile Device or Public / Shared Computer

When accessing your skills from a mobile phone, a guest laptop, or a public library machine, you don't need to leave your permanent account credentials linked. SkillVault lets you generate Temporary (Short-Lived) Read Tokens.

How Temporary Read Tokens Work

  1. Go to Token Settings (/dashboard/token).
  2. Choose an expiration window: 15 Minutes, 30 Minutes, 1 Hour, 24 Hours, or 7 Days.
  3. Click Generate Temporary Read URL.
  4. Copy the temporary URL (e.g. https://app.skillvault.dev/v/your-username/t/TEMP_TOKEN_STRING) and paste it into chat on your phone or shared browser.

AI chatbots on your mobile device or public computer can read and execute your skills for the duration you specified. Once the time expires, access is automatically revoked without leaving your account linked.

04 Uploading & Managing Skills via AI Agents (Publish Tokens)

Whenever you want an AI agent in chat or an agent harness on your computer to upload a new skill or modify an existing skill in your vault, you use a Publish Token.

🛡️ Why Uploading Requires a Publish Token

Uploading or overwriting skills can be a destructive action (a bad prompt could overwrite or delete existing skills). To keep your vault safe, SkillVault enforces a strict security rule: only short-lived Publish Tokens have write/delete authority. Your permanent read tokens cannot mutate your vault.

How to Have an Agent Upload a Skill

  1. Go to /dashboard/token and select a Publish Token lifetime (e.g. 30 minutes).
  2. Click Generate Publish Token and copy your Publish URL (https://app.skillvault.dev/v/your-username/p/PUBLISH_TOKEN_STRING).
  3. From a Desktop Agent Harness (Claude Code, OpenCode, Cursor):

    Tell your local agent in chat:

    "Upload the {skill name} skill to my SkillVault using this Publish URL: https://app.skillvault.dev/v/your-username/p/PUBLISH_TOKEN_STRING"

    The local harness will automatically package and push the skill along with all its supporting scripts, files, and frontmatter.

  4. From a Browser Chatbot (ChatGPT, Claude.ai):

    Attach any related script files directly in the chat session, paste your skill text, and give the agent your Publish URL.

05 Dashboard Smart Importer (Zip Archives & Folders)

If you prefer to add skills yourself directly from your web browser without talking to an AI agent, use the Dashboard Smart Importer at /dashboard/skills.

Uploading Files, Folders & Raw Prompts

  1. Navigate to your Skills Dashboard.
  2. At the top of the page, locate the Upload & Smart Importer card.
  3. Choose your input format:
    • Drag & Drop Zip Archives / Folders: Drop a .zip file or folder containing supporting scripts (Python, Node.js, bash), helper documents, or rule files (.cursorrules, CLAUDE.md, SKILL.md).
    • Paste Unformatted Prompt Text: Paste any system prompt, instruction block, or raw markdown snippet directly into the text area.
  4. Click Analyze & Normalize. SkillVault's AI engine automatically extracts a valid kebab-case name, 1-line description, standardized SKILL.md structure, and attached helper files.
  5. Review the preview, adjust any fields if desired, and click Save to SkillVault.

06 How Skill Execution Works & System Capabilities

SkillVault handles skill execution intelligently depending on how and where your AI assistant is running:

Skill Category On Paired Laptop / PC (CLI) On Web Chatbots, Mobile & Perplexity
Instruction-Only Skills
(Markdown guidance, rules, prompt standards)
Read directly from vault and followed inline by AI. Read directly from vault and followed inline by AI.
Skills with Code & Scripts
(Python scripts, Node tools, transcript extractors)
Executed locally using installed packages on your computer. Executed via SkillVault Cloud Proxy Engine.
Skills with Simple Auth
(API keys, SSH passwords, database tokens)
Substituted and executed locally via CLI. Executed via Cloud Proxy. Encrypted secrets are substituted in transit: the LLM never sees your keys!
Interactive / OAuth Skills
(Browser pop-ups requiring human clicks)
Runs on your local desktop machine where browser pop-ups occur. Must be run locally on desktop. Have your local agent download the skill locally first (via skillvault pull <name> or asking your agent to pull it into your workspace).

07 Storing API Keys & Passwords (Where & How)

Many skills require API keys or SSH credentials (such as an OpenRouter key, Apify key, or VPS password). SkillVault allows you to store credentials securely on the server so your AI agents can use them without you ever pasting secrets into chat windows.

How to Add Secrets to a Skill

  1. Go to your Skills Dashboard (/dashboard/skills).
  2. Find the skill that requires credentials (e.g. geo-kb or amz-nsr) and click the Manage button.
  3. Scroll down to the Stored Secrets card.
  4. Enter the Secret Name (e.g. vps_pass, openrouter_api_key, apify_key) and paste your Secret Value.
  5. Click Save Secret. Credentials are encrypted immediately using AES-256 encryption.
💡 Don't Know Which Secret Names Are Required?

Simply ask your AI agent in chat! Your agent reads the skill requirements from your vault and can tell you the exact secret names required. Once you save them in your dashboard, SkillVault handles substitution automatically!

08 Utilizing the Memories Tab (MemVault)

SkillVault does not just store instructions and tools: it provides MemVault, persistent cross-session cognitive memory for your AI coding agents. While skills teach your agents how to do things, MemVault ensures they remember what you decided across every terminal, IDE, and chat session.

How Memories Get Captured

You don't need to manually type notes into a database. MemVault captures context ambiently through three channels:

  • Automatic Git Commit Milestones: Whenever you commit code in a repository where you ran skillvault setup, the automatic git post-commit hook sends a rich milestone (commit message, diff summary, affected files, author, and branch) directly into your memory vault.
  • Proactive In-Chat Directives: Simply tell your AI assistant in Cursor, Claude Code, Windsurf, or ChatGPT:
    "Remember that we migrated our database ID format to UUIDv7"
    Your agent detects the milestone or preference and immediately calls the MemVault API or tool to store it.
  • Session Wrap & CLI Ingest: When concluding a major debugging session, tell your agent "Wrap up this session" to extract bug root-causes and architecture fixes. Or use the terminal CLI directly:
    TERMINAL · CLI MEMORY INGEST
    skillvault mem ingest --text="Always use Redis connection pooling in staging" --importance=high --project=backend

Tuning Agent Proactivity (CLAUDE.md & AGENTS.md)

Different developers and teams prefer different levels of agent autonomy. You have total sovereign control over whether your agent stays quiet until asked, or proactively queries MemVault before writing every line of code. Simply add the snippet matching your preferred style to your project's CLAUDE.md, AGENTS.md, or global ~/.claude/CLAUDE.md:

Option A: Passive / On-Demand (Default) ZERO CONFIG

Best if you want zero noise. The agent only recalls or stores memory when you explicitly ask, or when git commits fire automatically.

No file setup required. Simply say: "Check MemVault for our auth token rules" or "Save this decision to MemVault."
Option B: Proactive Pre-Flight (Recommended) MOST POPULAR

Your agent automatically checks team standards and past architectural decisions on Step 0 before planning or writing code.

CLAUDE.md / AGENTS.md / .cursorrules
# MemVault Directives - **Step 0 Memory Recall:** Before proposing architectural plans, refactoring, or choosing libraries, run `skillvault mem recall "<task or topic>"` (or use the `memvault_recall` tool) and strictly adhere to returned [ASKER CONTEXT] and pinned standards. - **Knowledge & Invariant Ingestion:** When establishing test/build recipes or non-negotiable constraints, run `skillvault mem ingest --text="Procedure: <exact command recipe>"` or `skillvault mem ingest --text="Invariant: <rule>" --pin`.
Option C: Continuous & Strict Guardrails STRICT COMPLIANCE

Enforces zero repeated bugs, mandatory recall on every turn, and automated session wrap summaries.

CLAUDE.md / AGENTS.md / .cursorrules
# MemVault Directives - Always run `skillvault mem recall "<current goal>"` at the start of any prompt. - When resolving tricky bugs, immediately document the root cause: `skillvault mem ingest --text="Trap: <root cause, platform quirk, and solution>"`. - At the end of every multi-step task, wrap the session with: `skillvault mem ingest --text="Decision: <key technical decisions made>"`.

Managing Your Memories in the Dashboard

Navigate to /dashboard/memories in your sidebar to visually audit, search, and curate what your agents know:

  • Live Instant Search: Type in the search box to filter memories in real time across content, author, project tags, and commit hashes.
  • Project Filter Tabs: MemVault automatically categorizes memories by project scope (e.g. [skillvault], [billing], [mobile-app]). Click any project tab to focus on that codebase.
  • Pinning Permanent Rules (📌 Pinned): Normal working memories naturally follow an Ebbinghaus forgetting curve if they aren't recalled over time. Clicking Pin locks that memory's strength at 1.0 forever, making it exempt from decay so it is always prioritized in prompt recall.
  • Editing Content & Metadata: Click Edit on any card to update the text, change its importance level, or reassign its project tag. MemVault automatically generates fresh high-dimensional vector embeddings so semantic search reflects your changes immediately.
  • Forgetting Outdated Decisions (🗑️ Forget): If a technical direction is reversed, an old architecture is deprecated, or a staging token was captured by accident, click Forget. MemVault uses a two-step confirmation and atomically purges the memory from vectors, associative knowledge graphs, and active recall.
🏢 Switching Between Personal & Team Vaults

If you belong to an organization, the Memories tab displays a Vault Switcher at the top:

[ 👤 Personal Vault ]   [ 🏢 Acme Corp (Team) ]

Your Personal Vault keeps local machine setups and scratchpad notes private. The Team Vault allows you to inspect shared team architecture decisions, bugfixes contributed by colleagues, and official Org Standards.

ORGANIZATION GOVERNANCE
Are you a Team Lead or Organization Admin?

Learn how to curate Org Standards, configure compliance auto-purge windows, and manage team members.

Read Team Admin Guide →

Quick Reference Guide

Scenario Recommended Method
Pair Laptop or PC Send Set up my SkillVault at https://app.skillvault.dev/v/username in chat & click Authorize.
Connect Web Account (ChatGPT, Claude, Perplexity) Add MCP URL or Custom API Connector from chat bar/menu. Configure once, access anywhere.
Mobile Phone or Shared Laptop Generate a 15-Min to 7-Day Temporary Read Token at /dashboard/token and paste link in chat.
Have Agent Upload a Skill Generate a 30-Minute Publish Token at /dashboard/token and give link to agent.
Upload Folders or Code Archives Drag & drop .zip or folder into the Smart Importer at /dashboard/skills.
Add Passwords & API Keys Go to /dashboard/skills → Manage → Stored Secrets. Never paste keys in chat!
Inspect & Search Agent Memories Visit /dashboard/memories to search, filter by project, pin rules, or forget outdated decisions.
Pin Permanent Rule or Architecture Click Pin on any memory card or run skillvault mem pin <id> to lock strength at 1.0 (no decay).
Manage Team & Org Governance Visit /dashboard/orgs and read the Team Admin Guide.

Enjoy your seamless, cross-device AI skill library!