Arca MCP Server

Step 1: Sign up for a free account at https://arca.build

For ChatGPT

1. Enable Developer Mode in ChatGPT settings
2. Go to Custom Connectors
3. Add a new connector with:
   • Name: Arca
   • URL: https://mcp.arca.build
   • Authentication: OAuth
   • Follow the OAuth flow to sign in with your Arca account

For Claude

1. Go to Custom Connectors in Claude
2. Add a new connector with:
   • Name: Arca
   • URL: https://mcp.arca.build
   • Authentication: OAuth
   • Complete the OAuth login using your Arca credentials

1. Get Your API Key

Sign in to Arca and go to your Account page. Copy your API Key.

2. Configure Your Agent

Connect to the MCP server using Bearer authentication:

URL: https://mcp.arca.build
Authorization: Bearer YOUR_API_KEY_HERE

Discovery Tools

• list_available_tools - See all available MCP tools
• list_skills - Call this first! See all your skills and their types

Creating Skills

• create_tabular_skill - Create structured data skills (todos, meals, workouts, check-ins)
• create_vector_skill - Create semantic memory skills (journal, preferences, reviews)

Working with Tabular Skills (Structured Data)

• add_tabular_item - Add new records with consistent fields
• get_tabular_items - Query with filters, sorting, and SQL-like aggregations
• update_tabular_items - Update existing records by ID or condition

Working with Vector Skills (Semantic Memory)

• add_vector_item - Store text entries for semantic search
• search_vectors - Find similar content by meaning, not exact match

Creating Your First Skill

You say: "Track my daily meals with calories and protein"

Triggers: create_tabular_skill

Creates a "meals" skill with fields like food_name, calories, protein, timestamp

Adding Data

You say: "I just ate a chicken salad with 450 calories and 35g protein"

Triggers: add_tabular_item

Adds the meal to your meals skill

Querying Data

You say: "How many calories have I eaten today?"

Triggers: get_tabular_items

Queries meals with date filter and sums calories

Storing Preferences

You say: "Remember that I love Thai food and prefer spicy dishes"

Triggers: create_vector_skill + add_vector_item

Creates/updates preferences skill and stores this as semantic memory

Semantic Search

You say: "What kind of Asian cuisine do I like?"

Triggers: search_vectors

Searches preferences for semantically similar content about Asian food

📝 Todo List Management

→ Creates tabular skill "todos" with task, priority, due_date, completed fields

→ Adds item with add_tabular_item

→ Queries with get_tabular_items (filter: completed=false AND priority=high)

💪 Workout Tracking

→ Creates "workouts" skill with exercise, sets, reps, weight fields

→ Logs workout with add_tabular_item

→ Queries and aggregates with get_tabular_items (filter by date range and exercise)

📔 Journal & Memories

→ Creates vector skill "journal_entries"

→ Stores entry with add_vector_item

→ Searches with search_vectors (semantic match for "Sarah" and "hang out")

🍿 Content Preferences

→ Creates/updates "preferences" vector skill

→ Adds more preferences with add_vector_item

→ Searches preferences to understand your taste profile

💰 Expense Tracking

→ Creates "expenses" skill with amount, category, description, date

→ Logs expense with add_tabular_item

→ Aggregates with get_tabular_items (SUM amount WHERE category='food' AND month=current)

📊 Habit Tracking

→ Creates "meditation_log" skill with duration, mood_before, mood_after, notes

→ Logs session with add_tabular_item

→ Queries recent entries to calculate streak

• Always start with list_skills - Before adding or querying data, ask "What skills do I have?" so the AI knows what's available
• Be specific with queries - "Show my workouts from last week" is better than "show workouts"
• Update records naturally - "Mark my 'finish report' task as done" triggers update_tabular_items
• Use semantic search for fuzzy matching - Vector search works great for "What restaurants did I like in Italy?" even if you wrote "ate at amazing trattoria in Rome"
• Mix structured + semantic - Track meal calories in a table, but store recipe ideas and food memories in vectors

🔒 Privacy & Security

Your API key is your authentication credential. Keep it secure and never share it publicly. All data is stored in your private, isolated Arca vault with encryption at rest. Only you and the AI assistants you explicitly connect can access your data.