MyAIforOne — Admin

Loading channels...

Loading spaces...

Loading organizations...

Loading activity...

AI Models

Choose the model your agents run on, sign in to Claude & ChatGPT, store provider API keys, and optionally bridge to local open-source models via Ollama.

⚙

Claude Accounts

Default Claude Account

Used for agents without a per-agent account override. Switch when credits run out — saves instantly.

Add Account

Account Name

The config directory is derived automatically from the account name — main uses ~/.claude, any other name uses ~/.claude-{name}.

◆

Default Claude Model

The model every Claude agent uses unless overridden per-agent

Platform-wide default model

Passed as --model to claude -p for every agent. A per-agent /model command still overrides this. Leave on default to let Claude Code pick.

⚡

Alternative Models (beyond Claude)

Master gate for every non-Claude provider — the Provider API Keys grid and the Ollama bridge below. When off, every agent runs on Claude regardless of pasted keys or executor overrides.

Enable Alternative Models (beyond Claude)

On (default): pasted provider keys (OpenAI, Gemini, Grok, Groq, Together, Mistral, Venice, OpenRouter) and the Ollama bridge are active, and agents can be set to run on those. Off: everything below is greyed out and every agent uses Claude regardless.

🔑

Provider API Keys

Keys for alternative model providers. The e.g. in parentheses shows one example executor value — the API key unlocks the whole provider, and you pick the specific model from the Platform Default Executor dropdown above or on a per-agent basis. Any model that provider offers is available.

Anthropic / Claude (e.g. anthropic:claude-sonnet-4-6 — any Claude model)

OpenAI (e.g. openai:gpt-5.5 — any GPT model)

xAI / Grok (e.g. grok:grok-4.3 — any Grok model)

Google Gemini (e.g. gemini:gemini-3.5-flash — any Gemini model)

Groq (e.g. groq:llama-3.3-70b-versatile — any Groq-hosted model)

Together (together:meta-llama/...)

Mistral (e.g. mistral:mistral-large-latest — any Mistral model)

Venice (venice:llama-3.3-70b, grok-4-20, claude-sonnet-4-6…)

OpenRouter (e.g. openrouter:openrouter/auto — any of 345+ models across providers)

Gateway to 345+ models across OpenAI, Anthropic, Google, Meta, Mistral, and more — one key, one bill. Try openrouter:openrouter/auto for OpenRouter's auto-routing, or browse openrouter.ai/models ↗ for the full catalog. Set upstream provider preferences (which host serves each model) on openrouter.ai/settings/preferences ↗.

▣

Multi-Model via Ollama

Optional: run agents on local open-source models. Gated by the master toggle above.

Enable Tool-Calling for Alternative Models

When enabled, alternative-model agents (Ollama / OpenAI-compat / Anthropic-direct / Gemini-direct) gain a bounded tool loop with MCP + sandboxed file built-ins (Read/Write/Glob/Grep). Default off keeps them text-only. The Claude path is byte-identical regardless of this flag.

Platform Default Executor

Ollama Base URL

Saves every AI Models tab — Claude default model, Codex, provider keys, and multi-model settings — together.

Credentials & Connections

Two ways to give agents access to an app, side by side: paste-in API keys (inherited by agents) and first-party OAuth account connections (tokens held by the broker).

🔒

Encryption Status

🔑

Shared Keys

Defaults inherited by every agent that uses the MCP. Stored encrypted at PersonalAgents/mcp-keys/<mcp>.env.enc.

👤

Per-Agent Overrides

Read-only summary. Edit a specific agent's keys from Agents → Agent Config → MCPs.

Organization & Users

Your organization is this gateway's tenant — the company everyone here belongs to. Users sign in with their own email and password; roles control what they can do.

🏢 Organization

…

👥 Users

Loading users...

Context Engine

Controls the personal data pipeline — the engine that pulls from sources (Gmail, Calendar, Slack, Notion) on schedule and fans the records out to your contexts.

📊 Status

⚡

Engine Master Switch

When on, every enabled source runs on its configured cron schedule. When off, no scheduled pulls fire — but you can still trigger a manual pull from this page or from any source's detail page. Per-source and per-context enable flags still apply.

▾ 📊 Engine Stats

Loading stats...

▾ ⏰ Active Schedule

Loading schedule...

▾ 📜 Recent Pulls

Loading recent pulls...

🔧 Controls

▶ Manual run

Manage Sources & Contexts →

Pull All runs every enabled source sequentially, ignoring schedule. To re-register cron tasks after editing a schedule outside the UI, use Reload from disk inside the Active Schedule section above.

⚙ Configuration

▾ 🔑 Google OAuth Client

Required so Gmail and Calendar sources can pull on your behalf. One client serves all Google sources across all Contexts.

client_id client_secret

▾ 🔗 LinkedIn OAuth Client

Required so the LinkedIn MCP can be authorized from the agent Configure surface. One client serves every attached LinkedIn account (multi-account supported).

client_id client_secret

▾ 🛡 Personal Contexts settings

Optional capabilities for the Personal Contexts pipeline (Gmail / Calendar / Slack / Notion).

Credential Encryption

Encrypt every Source's credentials.json at rest with AES-256-GCM (same keystore as MCP keys). Toggling triggers an in-place migration on next boot — reversible.

Cross-Context Search

Enable the markdown scanner that searches across every Context's data/. Surfaces a search bar on the Contexts tab and the search_contexts MCP tool for agents.

Platform Updates

Check for new versions of MyAIforOne.

⬆ Version Status

Installed version: --

Latest version: --