Powerloom's Agentic Consumption Layer Is Now Open: Inside the BDS Datamarket

In our last post, we shared what DSV Mainnet and BDS look like after 2 months in production.

Here's what it looks like on the consumption side — a Telegram alert that arrived minutes after setup:

🐳 Whale alert: WBTC → USDC  $245,000

Pool: 0x99ac8cA7087fA4A2A1FB6357269965A2014ABc17
Epoch: 24785719
✅ Verified on-chain
   cid: bafkrei...
   project: allTradesSnapshot:0x26c4...

Every alert carries a CID that maps back to DSV-finalized state on the Powerloom anchor chain. Snapshotters computed the data, validators finalized it, the agent surfaced it — provenance preserved end-to-end.

Why this matters

Agents need reliable data. Most data APIs are "trust the vendor" — you take what you're given and hope it's right. BDS is the opposite: every response includes a verification object the agent (or you) can check against the on-chain commitment. The same cid/epochId/projectId that ships in the API payload also lives in ProtocolState.maxSnapshotsCid on Powerloom mainnet. If the CID matches, the data was finalized by DSV consensus before it reached you. That is the difference between a normal API response and a verifiable BDS response.

A decentralized validator mesh, finalized data, and now an agentic consumption layer on top — that's what we're opening up today.

Get a free API key (2 credits, no wallet)

Two ways to get your free key — pick whichever fits:

Option 1 — Browser UI (no CLI required): go to BDS metering service signup page, enter your email and agent name, and click Continue.

Complete the Cloudflare Turnstile challenge on the verification page, accept the Terms of Service, and click Verify.

Your API key appears immediately. Copy it — it's shown only once in this flow.

Option 2 — CLI: bds-agent signup runs the same device flow from your terminal. See the full CLI signup walkthrough.

The sk_live_... key is universal — it works against the metered HTTP endpoints, the OpenClaw skill, the hosted MCP server, and any framework that speaks Bearer auth. Top up later with native $POWER (50 $POWER → 10 credits + 2 bonus = 12 total) or any supported plan token; the same key is preserved.

The sk_live_... key is universal — it works against the metered HTTP endpoints, the OpenClaw skill, the hosted MCP server, and any framework that speaks Bearer auth. Top up later with native $POWER (50 $POWER → 10 credits + 2 bonus = 12 total) or any supported plan token; the same key is preserved. (Browser signup details)

Pick your path

Same key. Same /mpp routes. Same balance. Same verification objects. Pick whichever fits your workflow — or use both.

Path A — 🦞 OpenClaw one-shot (free key)

The powerloom-bds-univ3 skill on ClawHub is where OpenClaw meets BDS. With the free sk_live_... from above, the entire setup is a single prompt — no wallet, no on-chain payment.

Paste the free-key one-shot prompt into OpenClaw. The agent's first response is a question:

Got it. Let me gather the required inputs first.

(a) API key: paste your sk_live_... from `bds-agent signup`.
(b) Telegram alerts: Do you want Telegram alerts for whale swaps?
    If yes, paste POWERLOOM_TELEGRAM_BOT_TOKEN and POWERLOOM_TELEGRAM_CHAT_ID.
    If no, type 'skip' and alerts will print to stdout only.

You paste the key, then either both Telegram values or skip. From there the agent installs the skill from ClawHub, wires the env into OpenClaw config, and creates a Whale Radar cron firing every 15 seconds via whale-cron.mjs. No wallet credentials. No private key. No on-chain payment.

The flagship output is the Telegram alert from the top of this post. Every alert carries the full verification block — cid, epochId, projectId — that anyone receiving the alert can verify on-chain.

The skill ships:

• Whale Radar (whale-cron.mjs) — bounded bds_mpp_snapshot_allTrades, USD-threshold alerts, default cron at 15s
• Token-Flow — multi-pool, token-keyed monitoring across all indexed pools for a given token
• Autonomous DeFi Analyst — LLM-generated market analysis with random verify_data_provenance probes embedded in operation
• Telegram + Discord sinks — both surface the verification block in every alert

Want more credits? Upgrade to a paid plan

The 2 free credits cover roughly 1440 epochs of metered data. When you want a bigger plan, two upgrade paths:

1. Top up the same key — bds-agent credits topup against your existing sk_live_.... Same key, more credits.
2. Run the wallet-funded one-shot — paste the 08-openclaw-one-shot.md prompt instead. Same input-gathering pattern (API key + Telegram dispatch in chat upfront), but the agent then runs scripts/signup-pay.mjs to broadcast a native $POWER transfer on chain 7869 and claim a fresh sk_live_... for the launch_10_pl_power_cgt plan (50 $POWER → 10 credits + 2 bonus = 12 total). Use a burner wallet only for the private key.

The hosted MCP server at https://bds-mcp.powerloom.io/sse works beyond OpenClaw too — Claude Code, Cursor, LangGraph, CrewAI, anything that supports SSE MCP transport with a Bearer header.

(OpenClaw & Hosted MCP docs | Quickstart Path A | Free-key prompt | Wallet-funded prompt)

Path B — 🐍 bds-agent CLI (Headless, Programmatic)

For headless orchestration, CI pipelines, custom YAML recipes, or any environment where you want the agent driving its own queries directly against the metered HTTP endpoints. The CLI is uv-installable and works on the same sk_live_... from the free signup above.

git clone https://github.com/powerloom/bds-agent-py.git
cd bds-agent-py && uv sync && uv tool install .
bds-agent config init

config init writes BDS defaults — resolver base URL, endpoint catalog, Powerloom chain RPC, ProtocolState and DataMarket contract addresses — to your profile.

Query in plain language (LLM-backed; works with Anthropic, OpenAI, Ollama, local GGUF, or Apple Intelligence):

bds-agent llm setup anthropic && bds-agent llm use anthropic
bds-agent query "Top 5 Uniswap V3 swaps by USD volume in the last hour" --execute

Generate a YAML recipe from a description:

bds-agent create "Alert when any single swap exceeds $100k"

Run a recipe (no LLM needed for run):

bds-agent run examples/dex-alerts.yaml --profile default

The repo ships pre-packaged recipes including examples/dex-alerts.yaml, which streams indexed DEX swaps and triggers on min_usd (default $50k) and volume_spike (3× rolling average over 10 epochs). Set verify: true in any recipe to enable on-chain CID verification per batch — the runner calls ProtocolState.maxSnapshotsCid and routes a rule: verification alert on mismatch.

More credits when you need them: bds-agent signup-pay --plan-id launch_10_pl_power_cgt --chain-id 7869 --token-symbol POWER runs the same autonomous on-chain payment flow (50 $POWER → 12 credits) against your funded wallet, and the new key replaces the free one in your profile. See the Pay-signup reference.

(Full operator guide | Quickstart Path B | GitHub)

Verification: trust, but verify

Every BDS response from a supported route includes a verification object:

{
  "epoch": 24785718,
  "snapshot": { "...trade data..." },
  "verification": {
    "cid": "bafkreib6h6au6hyv3hnx3wmbysu4lcoouah2yicst3nwnwndpcw2yxxrta",
    "epochId": 24785718,
    "projectId": "allTradesSnapshot:0x26c44e5CcEB7Fe69Cffc933838CF40286b2dc01a:mainnet-BDS_MAINNET_UNISWAPV3-ETH"
  }
}

Anyone can verify independently:

cast call 0x1d0e010Ff11b781CA1dE34BD25a0037203e25E2a \
  "maxSnapshotsCid(address,string,uint256)(string,uint8)" \
  0x26c44e5CcEB7Fe69Cffc933838CF40286b2dc01a \
  "allTradesSnapshot:0x26c44e5CcEB7Fe69Cffc933838CF40286b2dc01a:mainnet-BDS_MAINNET_UNISWAPV3-ETH" \
  24785718 \
  --rpc-url https://rpc-v2.powerloom.network

If the CID matches, the response was finalized by DSV consensus before it reached the agent. That's the contract every alert carries. Both bds-agent run --verify and the OpenClaw verify_data_provenance MCP tool wrap this same call. Alerts emitted by either path include the full verification block so downstream consumers can run the same check. (Verification in Agent Workflows)

The SKILL.md pattern

Both bds-agent-py and the OpenClaw skill publish a SKILL.md at their repository root:

curl -sL https://raw.githubusercontent.com/powerloom/bds-agent-py/main/SKILL.md
curl -sL https://raw.githubusercontent.com/powerloom/powerloom-bds-univ3/main/SKILL.md

A single flat markdown file with the full command surface, metering HTTP flow, environment variables, common mistakes, and canonical links. Any LLM-driven agent or orchestrator can fetch it at session start and bootstrap itself without parsing full documentation.

This is the agent-native interface pattern we want more data services to adopt: framework-neutral orchestration directives in one file any agent can ingest. If you're building a data API and you want autonomous agents to use it, publish a SKILL.md. (SKILL.md details →)

Get started

• Full Docs: Agents & BDS | ClawHub skill | GitHub: bds-agent-py | GitHub: powerloom-bds-univ3
• Metering & API Keys | Headless CLI Guide | Verification

What's next

• More recipes — price, volume, and flow-imbalance regime change alerts
• More data markets — Aave V3 next, same bds-agent and ClawHub patterns
• More integrations — OpenClaw was the first, not the last. Hermes, LangGraph, CrewAI, Claude Code, and Cursor are all in scope.

Walkthrough videos for both paths are dropping shortly.

Spread the word: Discord | X | Telegram | GitHub | Website