A CLI for bootstrapping AI agent workspaces with ready-to-run workflows.

Workflows are .prose programs that orchestrate multi-agent systems using OpenProse, AgentWallet, and the Frames Registry.

npx wordspace init

This will:

1. Show available workflows and let you pick which ones to download
2. Configure agent permissions
3. Create the output/ directory

Then run a workflow:

npx wordspace run workflows/x-daily-pulse.prose

Wordspace auto-detects which coding agents you have installed and lets you pick one.

interactive -- hands over the terminal for live interaction headless -- executes the prompt and exits passthrough -- outputs workflow instructions for the calling agent (OpenClaw is already running when it invokes wordspace)

Use --harness to skip the picker:

npx wordspace run --harness aider workflows/x-daily-pulse.prose
npx wordspace run --harness claude workflows/x-daily-pulse.prose

• At least one supported agent installed
• An AgentWallet account (for paid API workflows)
• Wallet funded with USDC (even $1 covers hundreds of API calls)

Bootstrap a new project. Presents an interactive picker to choose which workflows to download.

npx wordspace init           # interactive setup
npx wordspace init --force   # re-download existing workflows

Browse available workflows from the repository.

npx wordspace search          # list all
npx wordspace search pulse    # filter by keyword

Download specific workflows by name.

npx wordspace add x-daily-pulse
npx wordspace add x-daily-pulse --force   # overwrite existing

Run a .prose workflow via a coding agent.

npx wordspace run workflows/x-daily-pulse.prose                    # auto-detect agent
npx wordspace run --harness claude workflows/x-daily-pulse.prose   # use specific agent
npx wordspace run github:frames-engineering/wordspace/workflows/x-daily-pulse.prose  # remote

npx wordspace run workflows/x-daily-pulse.prose

You'll be prompted for:

• accounts -- X handles to monitor (e.g., elonmusk, openai)
• topics -- keywords to track (e.g., AI agents, LLM tooling)
• timeframe -- how far back to look (default: 24h)
• goal -- your engagement goal (e.g., grow my AI agents audience)

The workflow spawns four specialized agents (data-fetcher, analyst, engagement-coach, briefing-writer) that fetch live Twitter data, analyze engagement patterns, and produce a markdown briefing saved to ./output/.

Cost: ~$0.10-0.20 per run depending on the number of accounts and topics.

                    ┌──────────────┐
                    │  .prose file │
                    └──────┬───────┘
                           │
                    ┌──────▼───────┐
                    │ OpenProse VM │  (your agent becomes the interpreter)
                    └──────┬───────┘
                           │
              ┌────────────┼────────────┐
              │            │            │
        ┌─────▼─────┐ ┌───▼───┐ ┌─────▼─────┐
        │  Agent 1   │ │Agent 2│ │  Agent 3   │  (parallel sub-agents)
        └─────┬──────┘ └───┬───┘ └─────┬──────┘
              │            │            │
        ┌─────▼────────────▼────────────▼─────┐
        │         Frames Registry              │  (pay-per-call APIs)
        │    Twitter · Exa · AI Gen · Test     │
        └─────────────────┬────────────────────┘
                          │
                    ┌─────▼──────┐
                    │AgentWallet │  (x402 micropayments)
                    └────────────┘

1. OpenProse defines the workflow as a .prose program
2. Your coding agent becomes the VM and spawns parallel agents
3. Agents call external APIs through the Frames Registry
4. AgentWallet handles payments automatically via the x402 protocol
5. Results flow back through the agent pipeline into a final output

• OpenProse -- the programming language for AI sessions
• AgentWallet -- wallets for AI agents
• Frames Registry -- pay-per-call API gateway
• x402 Protocol -- HTTP payment standard