Swarm Network · field guide

Launch a swarmbefore lunch.

BuildYourFuture.app is Swarm Network’s browser gateway for multi-agent systems on Sui. No local Docker, no pile of vendor keys. You pick the agents, set the boundaries, and the runtime is already there.

HermesreasoningMercurytriggersOpenClawgatewayNova
~3 min
Deployment window
Versus 30–120 minutes for a local setup
Zero keys
No key sprawl
Tool traffic routes through an authenticated proxy
~$13M
License capital
Backing the Swarm Network ecosystem
$0.0012+
Pay-per-call tools
Metered utilities instead of extra SaaS seats

01How it works

Four steps. No environment to babysit.

The usual path is Python versions, a Docker daemon, a broker, and a key for every vendor. BYF keeps that stack on the other side of the browser.

  1. 01

    Initialize the workspace

    Open a browser and create the cluster environment. No local virtualenv, Docker daemon, or VPS to stand up first.

  2. 02

    Choose the architecture

    Pick baselines from the catalog — Hermes for reasoning, OpenClaw for external gateways, Mercury for low-latency checks.

  3. 03

    Define the boundaries

    Set objectives, prompts, and schema rules in the UI instead of hand-writing orchestration glue.

  4. 04

    Provision the runtime

    BYF wires the managed sandboxes, tool bindings, and state so the cluster can run without you babysitting processes.

Proxy, not a .env

Provider and tool keys stay off the machine running the agent.

Managed sandboxes

Containers and tool bindings come up with the cluster.

One ledger

Template unlocks and per-call tools bill in the same place.

Isolated memory

Worker scratchpads do not dump into every other agent.

02Why BYF

Local frameworks vs. a managed cluster.

DimensionLocal DIYBYF
ComputeA workstation, Docker, or a rented Linux node — and the dependency conflicts that come with them.Cloud execution containers provisioned from the browser.
CredentialsScattered provider keys in local env files. One leak and every vendor is exposed.No raw OpenAI, Anthropic, or tool keys sitting in local files. Calls go through a proxy.
BillingA stack of vendor subscriptions, minimum balances, and separate invoices.One ledger: a one-time template unlock plus sub-cent pay-per-call tools.
Time to liveOften 30 to 120 minutes of setup, debugging, and a first test run.A standardized window around three minutes.
Swarm shapeHand-rolled brokers, Redis queues, or a brittle custom state machine.Nova: sandboxed workers and a governed master workspace.

Minutes to a live agent

75 minutes is the midpoint of a 30–120 minute local setup. BYF’s window is about 3.

Up to ~96% less setup time

03The stack

Four roles. One workspace.

01

Nous Hermes

Reasoning and synthesis

Multi-step reasoning, task decomposition, and claim synthesis across long context.

02

OpenClaw

Gateway and scraping

Live web harvests, Discord and Telegram relays, and outbound webhooks.

03

Swarm Mercury

Low-latency triggers

High-frequency checks, threshold audits, and rule evaluations without extra model overhead.

04

Nova

Workspace federation

The fabric that isolates worker noise and commits only verified output to master memory.

Template unlock, once

Nova is free to configure. Specialized templates are $5–$8.

What a single call costs

External tools meter in fractions of a cent. No extra SaaS seat.

04Nova

Workers can be noisy. The master workspace should not be.

If every scratchpad thought and failed API retry is broadcast to the swarm, the context fills with junk and the bill follows. Nova keeps that work inside a sandbox and promotes only what passes the boundary.

Worker

Isolated sandboxes

  • Private scratchpads and trial reasoning
  • Raw scrapes and retry loops
  • That noise stays off the shared bus

Boundary

Validation gate

  • JSON schema checks
  • Directional read and write rules
  • Sanitize before anything is committed

Master

Shared workspace

  • Project state other agents can trust
  • Clean inputs for synthesis
  • An auditable final assembly
  1. Phase 1

    Decompose

    The mission is split into sub-tasks aimed at specialized agents.

  2. Phase 2

    Work in parallel

    Workers run at the same time, each inside a private scratchpad.

  3. Phase 3

    Filter

    The boundary strips intermediate junk and checks the schema.

  4. Phase 4

    Commit

    Only the validated payload is promoted into the master workspace.

  5. Phase 5

    Synthesize

    Downstream agents read that clean state and produce the deliverable.

05Operating cost

Price the calls, not a seat.

Drag the fleet and the daily call volume. Chainlink and Stableenrich are both $0.0024. The unlock figure is an estimate — $2.50 per agent — not a quote.

What a month might cost

Utility spend is the metered calls. Fleet size only changes the one-time template estimate ($2.50 per agent here — a blend of the $5–$8 unlocks, not a live invoice).

Monthly utilities

$21.60

One-time unlock estimate $10.00

4 agents
200/day
100/day
Chainlink · $0.0024 × 200/day × 30
$14.40
Stableenrich · $0.0024 × 100/day × 30
$7.20

06Workflows

Four ways a cluster actually earns its keep.

Research

01

Fact-checked research desk

OpenClaw pulls articles and papers. Oneshotagent ($0.0012/call) breaks them into claims. Hermes cross-checks those claims in the master workspace and scores confidence.

OpenClaw → Oneshotagent → Hermes

DeFi

02

Oracle telemetry sentinel

Mercury polls Chainlink feeds ($0.0024/call) inside its sandbox. If liquidity or volatility breaks a bound, it commits an alert and OpenClaw fires the notification.

Mercury → Chainlink → OpenClaw

B2B

03

Lead qualification

A worker sandbox calls Stableenrich ($0.0024/call) for executive profiles and tech stacks. Records that qualify move to Hermes for a pitch draft.

Ingest → Stableenrich → Hermes

Media

04

Multi-channel content studio

OpenClaw watches feeds. Promoted briefs fan out to writer agents in separate sandboxes. Mercury checks compliance before anything publishes.

OpenClaw → writers → Mercury

07Sui, Walrus, $TRUTH

Compute stays managed. Settlement does not.

Agent runtimes live in managed containers so they stay fast. Licenses, proofs, and the audit trail anchor to Swarm Network, Sui, and Walrus.

01

Sui settlement

Object-centric settlement for license checks, identity primitives, and low-latency state commitments.

02

Walrus archive

Deliberations, debate outputs, and consensus proofs can be stored for a public audit trail.

03

$TRUTH lease-back

License holders can lease operating rights back to active swarms, tying idle license capital to the verification economy.

Agent license

You need a license to run a swarm.

Swarm Network licenses are issued at license.swarmnetwork.ai. The link on this page includes referral code RICMOD from ambassador @WiseDegen. Same portal, same license — the code attributes the referral.

Get your license

license.swarmnetwork.ai