Agents

Pinata Agents are hosted OpenClaw instances. Each agent runs in an isolated container with its own environment, skills, and secrets. Pick a personality, connect an LLM provider, deploy — your agent is live.

​

Agents Workspace

Open Agents in the Pinata App sidebar to access the Agents workspace. The workspace has three tabs: My Agents, Skills, and Secrets.

​

Before You Start

Your agent needs an LLM provider API key to run. Add it as a secret before creating an agent — the setup wizard looks for it during the Connect step. Pinata supports three providers:

​

Creating an Agent

Click Create Agent to start the setup wizard. There are four steps.

​

1. Identity

Choose a personality template or create a custom one:

• Atlas — analytical, methodical
• Nova — creative, exploratory
• Sage — patient, accuracy-focused
• Custom — define your own personality and purpose

Give the agent a name and an optional personality description.

​

2. Agent Workspace

Select a workspace template. The Pinata Optimized Agent environment includes Node.js, Python, and common CLI tools. State persists and syncs automatically.

​

3. Connect

Choose your AI provider. If the API key is already in the Secrets vault, the provider shows “Key available.” You can also configure messaging channels on this step. Telegram Create a bot through @BotFather, copy the bot token, set a DM policy, and optionally restrict access to specific user IDs. Slack Create a Slack App at api.slack.com/apps, enable Socket Mode, and generate an App-Level Token with the connections:write scope. Under OAuth & Permissions, add these Bot Token Scopes: For public channel support, also add channels:history and channels:read. Install the app to your workspace, then paste the Bot Token and App Token.

​

4. Deploy

Click Deploy Agent. The container spins up and your agent starts running.

​

Managing Agents

Click into any agent to view its detail page with the Agent ID, version, and controls for Chat, Settings, and Restart. The detail page also includes:

• Snapshots — snapshots of the agent environment, recorded every minute when changes are detected
• Channels — Telegram and Slack configuration
• Skills — attached skill packages
• Secrets — environment variables injected at startup

​

Chat

Click Chat to open a direct session through the OpenClaw gateway. You can send messages, paste images, and start new sessions. The gateway dashboard also provides access to instance management, sessions, usage metrics, and logs.

​

Settings

Click Settings to open the full OpenClaw configuration panel covering environment, updates, authentication, channels, commands, hooks, skills, tools, gateway, models, and logging.

​

Skills

Skills are reusable capability packages pinned to private IPFS. You manage them at the workspace level and attach them to agents during creation. Each skill folder needs a SKILL.md with YAML frontmatter (name and description) and can optionally include a metadata JSON file for required environment variables. Click Upload Skill, name it (letters, numbers, hyphens, and underscores), add an optional description, and click Upload & Register. Once an agent is deployed, its skills are managed through the OpenClaw instance directly.

​

Secrets

The Secrets Vault stores encrypted API keys and credentials at the workspace level. You attach secrets to agents during creation, and they’re injected as environment variables at startup. Values can’t be viewed after saving. Click Add Secret, enter the environment variable name (e.g., ANTHROPIC_API_KEY) and its value, then save. Each secret shows which agents are using it. After deployment, per-agent secrets are managed through the OpenClaw instance