DYNAMO — THE AI AGENT FINANCE AND ECONOMIC LAYER

One budget.
Many agents.
Paid by the second.

The finance and economic layer for the AI agent economy. Meter any work by the second — tokens, seconds, bytes, calls — settle it per signed voucher, and get paid as it delivers. Sub-cent economics, no chargebacks, no billing stack to build.

x402-compatible · non-custodial on every rail · your payloads never leave your machine

dynamo — spending-cap session
09:02$ npx @dynamoprotocol/mcp monetize
09:02✓ your MCP tool is metered per call - any buyer can pay per use
09:04$ buyer: openBudget $50 · escrow-funded, one click
10:31call #1 - 120 units metered - voucher signed
10:31✓ settled: +$0.012 to you - as the work runs
10:58call #47 - settled: +$0.61 - receipt signed, verifiable offline
14:22✓ week total: $47.23 earned - no invoice, no chargeback, no billing code
# every unit signed. every settle final. 1% on the token rail.
14:23✓ and if anything ever runs wild on the buyer side - the spending cap fires. budget intact.

A simulated week: a builder monetizing an MCP tool — metered per call, settled per voucher, paid as work delivers.

0 wait
you're paid as the work runs, per signed voucher
99+% earn
stays with you, no marketplace cut
1 budget
covers every agent, tool, and API in a workflow
100%
of your work stays on your machine
10 lines
of code to your first metered, paid call

WHY DYNAMO

Programmable, Instant, Low-Fee Settlement

For developers building AI services, the urgent questions are the same everywhere: how do I charge per use, how do I get paid now, and how do I keep what I earn. Dynamo is the layer that answers all three — natively, for agents.

PAID AS WORK RUNS

Settlement inside the work, not after it

Money moves to you inside each settle — call by call, episode by episode. Not an invoice, not a payout schedule, not thirty days. Your earnings are signed, final, and yours the moment the work is delivered.

SUB-CENT ECONOMICS

Price anything, keep everything

Integer micro-USD metering with no fixed per-transaction tax: ≈1.32% all-in on a $0.50 job, versus 62.9% by card (2.9% + 30¢). Micro-pricing becomes a product feature instead of an impossibility — and the fee on the token rail is 1%, not a marketplace's 20–25%.

PROGRAMMABLE BY DESIGN

Payments as code, for agents

x402-compatible challenges, EIP-712 signed vouchers, scoped agent credentials, budgets that fan out across a whole network of services. Programmable payment logic on the token level — not a checkout page.

MULTI-AGENT BUDGETS

Multiparty Orchestrated Payments

Per-call rails, session rails, and escrow tools all reason about a single payer and a single payee. Real workflows fan out: an orchestrator hires a dozen services at once. Dynamo is built for that shape — one budget, many streams, all metered and capped together, refused at the money layer the instant a cap is reached.

ORCHESTRATOR

budget: $50 / day

Split once, enforce everywhere — per child, per branch, and across the whole network at once.

INFERENCE AGENT

cap $20 · $0.002/sec

SLA-gated — degraded units are never paid for.

DATA FEED

cap $15 · metered per second

Revocable mid-task, without touching the rest.

SUB-AGENT

cap $15 · HARD STOP at cap

Past the cap: HTTP 402. Nothing reaches upstream.

Split once, enforce everywhere

Carve a parent budget into per-agent allowances. Caps hold per child, per branch, and network-wide under concurrency.

Aggregate back into one session

Dozens of parallel streams reconcile into a single auditable record: what each agent consumed, when, what it cost — to the second.

Revoke mid-task

Cut any child stream instantly without touching the rest. The orchestrator stays in control of the fleet while work is running.

STREAMING SETTLEMENT

Metered Stream Payments, by the Second

Discrete payments are all-or-nothing. A metered stream reacts to reality second by second. Every control below settles on objective, verifiable facts — latency percentiles, error rates, availability, unit counts — never on anyone's opinion of the output.

VERIFIED VALUE

Pay for what's delivered

Tie payment to a live SLA. When the service drops below the agreed bar, the stream throttles or stops on its own. SLA-failed units are never paid for.

REFUNDS

Partial-delivery refunds

If a stream degrades halfway through, refund the seconds after it broke — not the whole thing, not nothing. Buyer made whole first, from held funds.

RATE ANOMALY

The runaway stop

A cap stops the total. Rate detection stops the runaway — the moment an agent burns faster than its authorized envelope, before the ceiling is even reached.

INTEGRITY

Mid-stream integrity

If a long-running stream drops, billing stops at the last settled second with a provable stop-point. Nothing stranded, nothing double-charged.

SHOWCASE — BUILT ON DYNAMO

Games, Videos, and More

The first businesses built on Dynamo: they get paid per delivered unit — and every job runs under a cap made of money, so a runaway pipeline can't spend past what the user paid. Both are code-complete applications running the escrow rail on testnet, with the same SDK and the same rules the docs describe.

WordsToGames
keywords in → playable browser game out

Type a few keywords; an LLM pipeline writes a schema-enforced design document, builds a self-contained Phaser 3 game, and a headless browser plays it for real — pressing keys, reading game state, verifying every promised mechanic — while a bounded, separately-capped repair loop fixes what fails.

Dynamo inside

  • The user's stablecoin payment escrows as the job's budget — the cap is the amount paid; design, codegen, and each repair round run as their own capped, metered streams.
  • The developer is paid as the work runs — money moves to the payee inside each settle, not on a payout schedule days later.
  • A diverging repair loop is stopped mid-flight by the rate envelope — while most of the budget is untouched.
  • Unspent budget releases and refunds the user; every billed call leaves a signed dynamo-receipt-v1, reconciled three ways against the vault's own records.
funding:"usdc" escrowBase Sepolia testnetdelegated repair subtreescode-complete
WordsToVideos
keywords in → a numbered video series out

Type up to ten keywords, pay in stablecoins, receive a series of ~30-second episodes: a series bible fixes characters, palette, and tone across every episode; each shot is rendered, then metered — a failed shot never costs the episode.

Dynamo inside

  • One capped, per-second-metered stream per episode; headroom checked before every paid shot — throttled clamps, checkpoint-required settles first, refuse stops the episode.
  • Billing follows delivery — a hung or failed provider call bills nothing; retries can't double-bill (idempotent per request).
  • Paid per episode, during the job — force-checkpoint before every revoke, and the receipt shows each call's signed voucher.
  • The Spending Savings Board shows the measured economics: settlement latency, cost split, and the card-rails comparison — ≈1.32% all-in vs 62.9% by card on a $0.50 job.
per-episode streamsBase Sepolia verifiede2e 43/43 greencode-complete

Both applications run on the escrow rail against Base Sepolia testnet today — real contracts, testnet stablecoins. Mainnet money rails activate per instance after external security review. Sources: codebase-reviewed developer descriptions of bytecardinal/WordsToGames and bytecardinal/WordsToVideos.

GET PAID

Start earning in three steps

Your own dedicated instance, provisioned by hand while we're early. You run only the open client — your code and your customers' payloads stay on your machines.

01

Request your instance

One form. A person stands up your isolated instance — own process, own database, own keys — and verifies it before you get the URL and tokens. takes a minute · instance within about a day

02

Meter your service

monetize() on an MCP tool, a gateway plugin on LiteLLM/Kong/APISIX, or the SDK in front of any API. Fine-grained metering by the second, the call, or any unit you deliver. one command, or a few lines

03

Get paid as work delivers

Per-voucher settlement to your account, a signed receipt for every unit, and nothing between you and your earnings but 1%. The same rail answers every buyer's budget — so you get paid by card-funded buyers when the card rail opens. the proof is the demo

Capping your own agents' spend instead? Same instance, one command: npx @dynamoprotocol/guard up — a hard spending cap in front of any OpenAI-compatible tool.

FUNDING MODES

Three ways to fund, one rulebook

Budgets, allowances, caps, hard stops, revocation, and signed vouchers behave identically across every funding mode. Moving from a free trial to real money is a one-line change — not a migration, not a different product.

CONTROL — LIVE TODAY

Build free, verify everything

The complete protocol with no payment setup: develop, test, and prove your integration end-to-end against the exact machinery the money rails run. Free permanently — and when you flip to a money rail, nothing about your code changes.

CARD — PER INSTANCE, AFTER REVIEW

Card-paying customers, no crypto needed

A card authorization backs the buyer's budget; you're paid by split transfer as work delivers, from buyers who never touch a wallet. Money stays inside a licensed, audited payment institution — Dynamo never holds funds, never sees card numbers.

STABLECOIN — PER INSTANCE, AFTER REVIEW

Escrow that pays builders

Self-custodied contract escrow: sub-cent metering, 1% on the token rail, no chargebacks, no admin keys over anyone's funds. Live on Base Sepolia testnet today — mainnet activates after external security review.

Non-custodial on every rail: Dynamo never takes possession of funds, never operates stored-value balances, never converts between fiat and crypto. Card and stablecoin rails are enabled per instance, after external security review — never by default.

USE CASES

Where per-second beats per-call

Start with the agent economy — the buyer building today. The same streaming engine reaches much further.

COMPUTE & INFERENCE

Bill the seconds consumed

GPU time and token-by-token inference with a hard ceiling and an SLA gate per job.

REAL-TIME DATA

Pay while subscribed

An agent on a live market or sensor feed pays while subscribed — and stops the instant it unsubscribes.

AGENT-TO-AGENT

Agents hiring agents

One agent hires several others for a long task and streams payment against scoped, capped, revocable allowances.

METERED APIS

Rent by the second

APIs, data, and compute by the second instead of a monthly tier nobody fully uses.

VIDEO

Pay per second watched

Watch 90 seconds, pay for 90 seconds. And because settlement ties to objective delivery facts, a feed that buffers bills less.

GAMING

Play — and get paid — by the second

Pay-as-you-play access, instant in-game purchases, and payouts flowing the other way: one pot split across players, creators, and platform.

AND FURTHER

Any delivered unit

Live audio, music, tipping, bandwidth, energy — anywhere value should flow only while it's being delivered.

PROOF, NOT PROMISES

Verify us instead of trusting us

OPEN CLIENT

Read the code

The client, plugins, and conformance kit are Apache-2.0 and public. The boundary between open and closed is machine-verified on the exact published bytes.

github.com/DynamoProtocol/dynamo-open →

CONFORMANCE KIT

Run the checks on us

A public 12-check suite verifies any implementation — including ours — and emits a signed report. npx @dynamoprotocol/conformance-kit

DOCS THAT RUN

Every sample executes

Each code sample in the docs runs in CI against a live core. What you read is what runs — the build fails before a rotten sample reaches you.

docs.dynamo.zone →

RUNNABLE PROOFS

The receipts, not the promises

A recreated $47,000 runaway loop stopped at $63 · an evidence bundle verified offline through 71 checks, with no network · 20+ property suites at 10,000 randomized runs each · decision parity across four plugins in three languages.

FAQ

Questions, answered

What makes this different from x402 or session-based rails?+

They move money one payer to one payee. Dynamo settles payments across a network of agents — one budget split across many, metered and capped together, reconciled into one session. It stays x402-compatible, so it adds the multi-agent and streaming layer rather than replacing the standard.

Do I need crypto to use this?+

No. Every instance runs Control Mode today: the complete enforcement engine — caps, rate envelopes, anomaly stops, revocation, signed vouchers — with no money anywhere and no payment setup of any kind. Card and stablecoin rails are enabled per instance when you need them. The objects are identical across all three, so moving to a money rail is a one-line change rather than a migration.

What if a service underdelivers after it's been paid?+

Settlement is tied to objective facts — latency percentiles, error rates, availability, unit counts — signed and attested, never a subjective judgement of output quality. SLA-failed units are never paid for, degraded streams refund to the attested break-point with the buyer made whole first, and a declared holdback window separates captured from final. Recourse is built into the rail rather than promised in a policy.

Does anything I send reach Dynamo?+

No. The guard and the SDK run on your machine and verify credentials offline; what crosses the wire to your instance is metering facts only — stream ids, unit counts, amounts, signatures. Prompts, completions, and payloads never leave your environment. Real money, when you enable a money rail, moves only inside a licensed payment institution or a self-custodied on-chain escrow — never through Dynamo.

Does this only work for AI agents?+

Agents are the first buyer, but the engine is general. The same per-second streaming and budget-splitting powers pay-per-second video, pay-as-you-play gaming with player payouts, live audio, and metered human-facing APIs.

One budget. Many agents.
Paid by the second.

The finance and economic layer for the agent economy — built for the builders of it.

Request your instance See the demo View on GitHub