Skip to content

reference/stable

Roadmap

Where Pulse is, what ships next, the conditions each stage must meet, and how progress is measured.

Roadmap

Pulse ships in stages. Each stage has a purpose, a set of deliverables, the condition it must meet before the next one starts, and a number we publish to show whether it is working. Dates are targets and are revised here, with a changelog at the bottom, whenever they move. Nothing on this page is a promise of a feature, a date, or a token outcome.

Status: Stage 1 — Operators and apps · last updated 2026-09-20

Stage 0 — Standard and scoreboardlive

Purpose. Make agent activity on Robinhood Chain visible with an open, zero-cost label, and publish the numbers live.

Delivered

  • PULSE-1, the 24-byte calldata label: format, placement (direct and ERC-4337), safety rules, framework registry by pull request.
  • Indexer at chain head with conservative attribution (bound / unclaimed / test), reorg detection, reconciliation.
  • Public scoreboard and feed with the claimed-only default view and the operator count.
  • TypeScript and Python SDKs with fork-tested safety (tagged transactions are proven identical to untagged for every default-safe call type).
  • MCP server for Claude, Cursor and other assistants; llms.txt and llms-full.txt for the chain.
  • Public read-only API with OpenAPI, SSE streams, and /health.

Published numbers. Agents active (claimed and all-labelled) 0 claimed / 0 labelled, operators 0, agent transactions 0, USDG moved $0.00, frameworks 0, indexer lag 0 blocks.

Exit condition (met). A labelled transaction from an external adopter appears on the feed within one block; scoreboard, API and docs live on agentpulse.fi.

Stage 1 — Operators and appsin progress target: two weeks

Purpose. Give the people who run agents a reason to claim them, and give apps a way to treat labelled agents differently from bots.

Deliverables

  • Claim flow: wallet-signed claims, sender binding, agent cards with name, framework and history.
  • Operator dashboard: per-agent spend, failures, gas, dapps used, test traffic, unclaimed-use warnings.
  • Alerts: failure rate, daily spend, unknown sender using an ID, inactivity; webhook and email delivery.
  • Agent lane: request signing in the SDKs and requirePulse middleware for Hono and Express, with a reference merchant.
  • Badge: "used by N agents this week" for verified dapp owners.
  • Framework and dapp leaderboards with 30-day series; /verify for any ID or transaction.
  • Monthly State of Agents on Robinhood Chain report, generated from the data and published on the site.

Published numbers added. Claimed agents 0, operators 0, apps offering an agent lane, badge installs.

Exit condition. At least three operators have claimed agents from at least two frameworks; at least one app enforces the agent lane in production; the first monthly report is published.

Dependencies. None external. Independent review (three mandates: adversary, abuser, operator) passes with zero critical and zero high findings before this stage closes.

Stage 2 — $PULSE and verified labelsplanned target: four to six weeks

Purpose. Make a verified label costly to fake, so agent counts and agent lanes can be trusted at scale.

Deliverables

  • PulseStake: stake $PULSE behind an agent ID; tiers; 7-day unstake cooldown; slashing by an arbiter behind an interface; suspension until re-staked. Balance-delta checks; owner is a 48-hour timelock.
  • Verified badge on the scoreboard; staked:n tiers in /verify and in the agent-lane middleware.
  • PulseDistributor: sponsor-funded campaigns paid nightly from indexed activity, with a public payout ledger and Merkle claims; unclaimed funds return to the sponsor.
  • Docs token page moves from Planned to Live, stating exactly what the token does and does not do.

Published numbers added. Staked agents, total staked, campaigns funded, payouts distributed, slashes.

Exit condition. Contracts pass the independent review and an external audit; AuditToken passes against the real token on a fork; testnet run with staked agents and one funded campaign; then mainnet on the operator's explicit decision.

Dependencies. The token exists on PONS (launched separately; Pulse does not issue it). Arbiter multisig and treasury addresses set. External audit scheduled.

What the token is not. Not a payment method on Pulse; not required to be counted, to use the API, SDKs or dashboard; not a revenue share.

Stage 3 — Ecosystemplanned target: quarter following Stage 2

Purpose. Make PULSE-1 the default way agents identify themselves on Robinhood Chain, and make the data useful beyond Pulse.

Deliverables

  • Framework integrations: official adapters for the most-used agent frameworks, submitted upstream where possible; framework codes assigned by pull request.
  • Wallet and explorer integrations: an "agent" mark on labelled transactions in at least one explorer or wallet, using the public parser.
  • Data access: bulk exports and a documented dataset schema for researchers; historical series API.
  • Governance: staked $PULSE votes on tier thresholds, slash ratios, admissible evidence, framework-code policy and label format changes; Snapshot first, on-chain later.
  • Dispute process for slashing with published rulings; a staked juror pool replaces the multisig arbiter.

Published numbers added. Frameworks with adapters, governance participation, disputes and rulings.

Exit condition. A second indexer run by a party other than Pulse reproduces the scoreboard from the public parser and the chain; governance has decided at least one parameter change.

Stage 4 — Sustainplanned

Operational commitments once Stage 3 is met: uptime and lag targets published on /health; a security contact and bug bounty for the contracts and the indexer; quarterly reports; a deprecation policy for label versions (a new version is announced at least 90 days before the old one stops being counted).

Not planned

  • Trading, portfolio or memecoin analytics features.
  • Custody of any funds or keys.
  • Charging in $PULSE for anything; fees, where they exist, are in USDG.
  • Counting unlabelled transactions as agents by inference.
  • Collecting identity from operators beyond a wallet and an optional name.

Risks we are watching

  • Sybil counts before staking. One operator can register many IDs; the scoreboard defaults to claimed agents and shows operators for that reason.
  • Standard adoption. A label nobody adopts measures nothing; Stage 1 and 3 are about adoption, and the framework count is the number that says whether it is working.
  • Chain and RPC dependence. Indexer lag and gaps are published on /health and the live dot; a dedicated archive endpoint and a second indexer are the mitigations.
  • Token expectations. The token has no function until the stake contract is live; this page and the token page say so at every stage.

Changelog

  • 2026-09-20Roadmap as a scorecard: derived chips, live numbers, changelog from git0776f10
  • 2026-09-20Phase 3 contracts: staking and campaign payouts, prepared but not deployedcd7eb24
  • 2026-09-20Public roadmap with exit conditions, and a "not planned" list87a8595