Help Center

Table of Contents

What is Braintrade?

Braintrade is a package registry and workflow toolchain for autonomous AI agents. It helps you export an OpenClaw agent workspace into a deterministic `.brainpkg`, publish versioned releases, discover packages, review them, and install them to connected agents.

Which features are available?

• Package export/import/publish/pull with integrity checks.
• Search, discover, ratings, comments, replies, and library saves.
• Agent registration, remote install jobs, and install status tracking.
• Merge workflows (Mix Agent) and version diff tools.
• Profile metadata editing, cover images, legal acceptance, and billing plans.
• CLI diagnostics (`braintrade doctor`) and dashboard diagnostics.

How does it work overall?

1. Build your OpenClaw agent workspace.
2. Export to `.brainpkg` with manifest + payload metadata.
3. Publish to Braintrade registry with semantic versioning.
4. Browse or search packages in web or CLI.
5. Install a selected version to one or more connected agents.
6. Iterate with review feedback, diff, and merge workflows.

How do I install and use the CLI?

Install commands:

# Linux / macOS
curl -fsSL https://braintrade.org/install.sh | sh

# Windows PowerShell
irm https://braintrade.org/install.ps1 | iex

Typical first commands:

braintrade doctor
braintrade auth login --device --output json
braintrade export ./my-agent --output ./my-agent-0.1.0.brainpkg --name my-agent --version 0.1.0 --description "demo"
braintrade publish ./my-agent-0.1.0.brainpkg
braintrade search my-agent

How do I install via Telegram message to my OpenClaw agent?

Braintrade does not require a specific Telegram bot implementation. The common setup is a Telegram bridge that translates a chat command into `braintrade` CLI execution on your agent host.

Recommended flow:

1. Register your agent in Braintrade and keep its connector token secret.
2. Run your Telegram bot only for authorized chat IDs.
3. Map a command like `/install my-agent 0.2.0` to a controlled local script.
4. That script runs `braintrade agent install ...` with fixed safety defaults.

# Example command your bridge can run:
braintrade agent install my-agent 0.2.0 --agent 1 --backup local --restart ask --output json

How do I use the web interface?

• `Browse`: discover packages, open details, check versions, ratings, and metadata.
• `My Agents`: manage your own packages, run Mix/Diff/Doctor, and restore deleted packages.
• `Connected Agents`: register agents and create rollout install jobs.
• `Profile`: legal status, provider links, subscription state, and plan controls.
• `Tokens`: create/revoke PATs for CLI and automation.

What files are in a typical OpenClaw package?

`.brainpkg` is a deterministic ZIP archive with this shape:

manifest.yaml
payload/...
assets/... (optional)

• `manifest.yaml`: package identity, engine compatibility, and file hashes/sizes.
• `payload/...`: the actual agent files used at runtime.
• `assets/...`: optional non-runtime assets (docs/media/etc.).

Common OpenClaw package content patterns include prompt files, tool configs, policies, knowledge files, and workflow scripts. The exact file names are your design decision, but the manifest must describe every payload file.

What makes a good specialized agent?

• Narrow scope: one domain, one success definition, clear boundaries.
• Deterministic inputs/outputs and explicit task contracts.
• Safety guardrails for destructive actions and external calls.
• Versioned prompt/tool changes with measurable regression checks.
• Operational observability: logs, failure states, and rollback strategy.

How do I design a strong OpenClaw workflow?

1. Define the target outcome and hard constraints.
2. Split workflow into deterministic stages with clear handoffs.
3. Attach a minimal toolset per stage; avoid broad tool exposure.
4. Add fallback paths for incomplete context or upstream failures.
5. Package, publish, test in staging, then roll out gradually.
6. Track quality signals and iterate using diff + merge workflows.

Security and safety checklist

• Do not include secrets in packages; use environment variables/secret stores.
• Review `manifest.yaml` and hash verification before installation.
• Use least-privilege tokens and revoke unused PATs.
• Run backups before installs on critical agents.
• Use staged rollout and verify health after deployment.
• Keep agent host and dependencies patched.

Troubleshooting

• Run `braintrade doctor` first to validate connectivity and auth.
• If web actions fail, capture the Request ID shown in error messages.
• For install issues, inspect install job status and agent logs.
• For TLS/domain issues, verify DNS, cert status, and Caddy logs.
• For quota failures, check plan/storage usage in Billing.

FAQ