Ponder docs

What Ponder is

Ponder is a private, verifiable AI agent on Robinhood Chain. You give it a goal; it plans the steps, does the research, checks its own work and hands back the finished result. Every task runs inside a hardware enclave that your own browser verifies before a single word leaves your device.

Where a chat answers one question at a time, Ponder carries a task from start to finish: long research, multi-step work, and jobs that keep running on a schedule while you are away.

Private thinking. Verified. That is the whole idea: the most capable models, working for you, with proof of where every step ran.

Quickstart

1. 1
   Open the agent

   Go to the Agent page. There is no account to make and nothing to install.

2. 2
   Pick a model and a mode

   Tap the model logo to choose from the top models, and switch between Agent mode (a task to carry out) and Chat mode (a question to answer).

3. 3
   Hand over the task

   Describe the goal in plain words. In Agent mode Ponder replies with a plan, works through it step by step, and ends with the result and next actions.

4. 4
   Keep the receipt

   Every run closes with a signed receipt you can share with anyone who needs to trust the result.

Agent mode and Chat mode

Agent mode is for work. Ponder writes a short numbered plan, carries out each step under its own heading, and finishes with a Result section and a list of next actions. Use it for research, comparisons, launch plans, reports and anything with more than one step.

Chat mode is for questions. Ponder answers directly, with structure when it helps, and keeps the conversation going for as long as you like.

Autonomous research

Name a topic and a format. Ponder searches widely, reads every source, cross-checks each claim against the others and writes a brief with citations you can follow.

• Dozens of sources per brief
• Every claim checked against at least two sources
• Citations inline, ready to share

Plans and checkpoints

Every task starts as a plan. You can approve it as it is, reorder steps, add one, or remove one before anything runs. Long tasks pause at checkpoints so you stay in control of the direction.

Scheduled agents

Any task can run on a schedule: every morning, every Friday, or at the first of the month. The result waits for you, sealed, the next time you open Ponder.

Every day · 08:00
Brief me on my watchlist: price, volume, news, and anything unusual.

On-chain watchers

Watchers keep an eye on Robinhood Chain for you: a wallet, a contract, a price or a pool. Write the rule in plain words and Ponder tells you the moment it fires. Watchers read the chain and never hold a key.

Models and routing

Ponder runs on the top models from every major lab and routes each step to the model that does it best: one model to plan, another to research, another to write.

Run receipts

Each run ends with a receipt signed inside the enclave. It names the enclave and its measured build, the models that ran, and a hash of every step and its output.

Share the receipt with a teammate, a client or your community, and anyone can confirm the work ran where Ponder says it ran.

Verification

Before your first word leaves the device, your browser checks the enclave's remote attestation: the hardware signature from AMD or NVIDIA, and a measurement that matches Ponder's open-source build. Only then does the task go out.

Encryption

Each task is sealed with Hybrid Public Key Encryption (HPKE, RFC 9180) to a key that only the verified enclave holds. The answer comes back the same way and opens on your device.

Your data

• No account, no email, no phone number
• Runs and results stay in your browser
• Backups sealed under your own passkey
• Zero trackers, and nothing is ever used for training

Daily holder credit

Every 1M $PONDER in your wallet earns one holder voucher of credit a day, starting at $0.40 each and able to rise with the market. Holding starts at 100K, which earns a voucher every 10 days.

Claim with one signature from your wallet, with no gas. Your browser blinds the notes before they are signed, so spending them never points back to your wallet.

Paying with $PONDER

Top up with $PONDER and every payment goes 20–25% further. Holders also get up to 30% off every top-up: 10% from 100K, 20% from 1M and 30% from 10M. The two stack.

Buyback and burn

$PONDER paid for credit feeds the burn, and a share of Ponder's revenue buys $PONDER on the market and burns it. The more Ponder is used, the smaller the supply becomes.

1. 1
   Use

   People run tasks and top up credit.

2. 2
   Pay

   Top-ups in $PONDER feed the burn directly.

3. 3
   Buy back

   Revenue buys $PONDER on the market.

4. 4
   Burn

   Bought-back tokens leave the supply for good.

API

The Ponder API is OpenAI-compatible. Point your client at Ponder, and the SDK verifies the enclave before each call; the rest of your code stays the same.

import { Ponder } from "@ponder/sdk";

const ai = new Ponder({ verify: true });
const run = await ai.agent.run({
  model: "kimi-k3.1-pro",
  task: "Compare the top 5 RWA tokens and write a brief",
});
console.log(run.result, run.receipt);