Build source-grounded crypto market Data Intelligence products on top of
TradeOS public evidence.
TradeOS is the crypto-market vertical of a broader source-grounded Data
Intelligence OS strategy. It is not a trading platform, broker, custodian,
managed account service, or hosted execution system. See the
TradeOS Business Thesis for the business
overview.
This repo gives builders SDKs, MCP tools, a BYOK agent CLI, runnable bots, and
the blueprint for the flagship TradeOS app pattern: a private self-hosted
Symbol Cockpit.
TradeOS hosts the intelligence API. The cockpit is an individual private-use
control plane: it runs on the self-hosted operator's own machine, server, quant
workstation, or agent host. It can recommend buy, sell, trim, avoid, watch, or
pass decisions from TradeOS evidence plus local rules, while keys, custody,
portfolio context, approvals, and execution stay in that operator's environment.
Builders can package or extend this pattern for users to self-host. They should
not operate the cockpit as a managed account service, collect customer exchange
credentials, custody customer assets, or place orders for customers.
That private self-hosted model is the product advantage. Traders get an
actionable agent. Builders get a product they can ship. TradeOS stays the
source-grounded intelligence product layer instead of becoming a custodian,
broker, or hosted order router.
Think of TradeOS as a crypto market Data Intelligence OS:
TradeOS supplies the source-backed intelligence layer.
Builders package that intelligence into briefings, bots, watchlists,
dashboards, validation packs, proof pages, and widgets.
Users and builder apps send structured feedback that improves intelligence
quality and provenance.
Paid TradeOS, x402, or enterprise access starts when a product needs scale,
alerts, automation, premium history, private intelligence products,
validation APIs, or explicit data rights.
The practical loop is simple:
Use TradeOS free.
Earn Data Intel Credits by improving intelligence quality.
Build and earn on public intelligence.
Pay when you need private intelligence products, scale, alerts, automation, or data rights.
Use the free public kit to prove a workflow. Charge for the packaging around
that workflow. Submit feedback with provenance so TradeOS can improve the
intelligence layer. Upgrade to paid TradeOS when the product needs production
volume, alerts, automation, premium history, private intelligence products,
validation APIs, x402 machine access, explicit data rights, or enterprise
support.
The flywheel effect is the point: useful TradeOS evidence helps builders ship
paid services; paid services create users, usage, feedback, and outcome labels;
that feedback improves TradeOS intelligence; better intelligence makes builder
products more valuable; successful products then need paid TradeOS depth, x402
access, and enterprise support.
Data Intelligence Services / SKU Map
TradeOS packages intelligence as product layers, not only as raw datasets:
Product Layer
Builder-Facing Shape
Access Model
Public intelligence products
Market Pulse, Platform Pulse, Token Radar, Fusion Signal Lite, Ask TradeOS, watchlists, and dossiers
Free public baseline with bounded read-only context
DTI credit unlock SKUs
Refresh boost, symbol pack, history window, evidence depth, Token Discovery detail, Fusion Signal detail, Review Lab passes, and AskTradeOS question packs
Human Data Intel Credits for temporary public GUI depth, public Ask packs, or read-only Review Lab access
Private intelligence passes
private_30m, private_1h, and private_4h private dashboard windows
Exact x402 pay-per-view sessions for private intelligence surfaces
Agent/API/x402 data SKUs
Agent router, Venice raw inference, Venice intel router, token risk, token discovery, risk-gated discovery, signal quality, signal evidence packs, fusion snapshots/history, Market Pulse Pro/team, VPIN stress, dataset concierge, and package scoping
Paid or reviewed production access for builders, agents, marketplaces, and API consumers
NEAR Intents-adjacent x402 wrappers
risk preflight, route context, asset dossier, status monitor, liquidity caveats, solver readiness, and dataset scope
Same read-only SKU family packaged for intent-aware apps; x402 pays for intelligence only, while NEAR Intents or the partner app handles execution
Keep the boundary explicit in builder products. Human DTI credits do not unlock
paid API scale, x402 calls, alert delivery, exports, automation, custody,
execution, private context, or enterprise data rights. Those are paid, reviewed,
or contract-entitled paths.
no TradeOS account required, server-side limits apply
Saved watchlists
require a signed-in TradeOS account token
Builder attribution
optional TRADEOS_PUBLIC_INTEL_KEY
DTI/API lifecycle
Data Intel Credit Loop explains human DTI, app reputation DTI, and feedback status endpoints
LLM inference
recommended BYOK default: Venice AI with e2ee-glm-5-1 for privacy-enhanced self-hosted inference; TradeOS also publishes paid AntSeed/x402 Venice raw-inference and evidence-backed intel-router SKUs
Paid machine access
x402 payment or TradeOS entitlement
Live platform proof
Platform Pulse shows feedback signals, x402 challenge demand, source mix, and settlement health
actionable recommendations are allowed; TradeOS does not custody keys, manage accounts, place orders, or support third-party account management
What This Is
TradeOS sits in the public market-intelligence layer for crypto and on-chain
markets. It is one vertical, not the entire Source Intelligence Network
strategy. This kit is the builder distribution surface for that Data
Intelligence OS. It helps products discover, consume, package, and send
feedback on current, source-backed market context inside bots, dashboards,
agent workflows, research products, validation systems, and paid communities.
It is an evidence and feedback layer:
public digest inputs, candidates, watchlists, thesis records, proof lookups,
caveats, freshness, source refs, and invalidation notes;
a public Fusion Signal Cockpit preview that shows redacted symbol-level
direction, confidence, freshness, why-fired context, invalidation, and
feedback intake without exposing execution fields;
symbol-cockpit and action-agent patterns for turning evidence into
good/bad/ugly verdicts, watchlist recommendations, and bot preflight checks;
non-executable action intents that carry evidence-backed action context into
local review, policy gates, paper execution, or independent executor
experiments without becoming orders;
TypeScript and Python SDKs for apps and services;
a stdio MCP server for local agent hosts such as Claude Desktop and Cursor;
a BYOK CLI that can ask Venice AI or another OpenAI-compatible model over
TradeOS evidence;
a market briefing bot that can post to stdout, Discord, or Telegram;
structured feedback writes with provenance so TradeOS can learn which public
intelligence was useful, early, late, thin, or confusing.
It is not a TradeOS-hosted broker, custody product, managed account service, or
order router. Builders can wire the open-source kit into local bot, preflight,
or execution stacks, but exchange keys, approvals, sizing rules, and custody
stay with the self-hosted operator using it for their own account context. If a
builder collects customer credentials, controls customer accounts, or runs
execution for other people, that builder has created a separate managed trading
service outside this kit's supported boundary.
Action intents are the bridge, not the executor. They are marked
non_executable, require operator review, and deliberately omit venue, account,
size, order type, route, calldata, transaction body, and execute URL fields.
In practical terms:
TradeOS Data Intelligence OS -> intelligence, evidence, data, feedback IDs
Symbol Cockpit -> private local recommendation and decision runtime
Local modules -> feasibility, EA/risk, execution adapters, ops dashboard
Why Builders Use It
Avoid weak intelligence gaps: reduce wrong-token routing, stale context,
unsupported claims, caveat-free recommendations, missed watchlist changes,
and outputs that cannot be audited. See Problem Space
for exposure anchors that translate weak context into potential dollar impact
without promising guaranteed loss prevention.
Understand the business thesis: TradeOS is the crypto-market Data
Intelligence OS vertical. It creates demand with useful public intelligence
and monetizes paid depth when workflows need scale, private intelligence,
alerts, validation, automation-safe access, or data rights. See
TradeOS Business Thesis.
Ship faster: start with bounded TradeOS evidence instead of building a
market data pipeline first.
Ground agents: give LLMs current evidence, caveats, stable IDs, and source
references before they answer.
Use privacy-enhanced BYOK inference: use Venice AI
as the recommended default with e2ee-glm-5-1, an E2EE/TEE-backed upstream
model option. Builders can still swap in another OpenAI-compatible provider.
Builders that want a marketplace-hosted path can use paid TradeOS
AntSeed/x402 SKUs:
tradeos-venice-raw-inference for ungrounded model output, or
tradeos-venice-intel-router for TradeOS evidence-backed model summaries.
Earn on top of TradeOS: package TradeOS intelligence into services,
workflows, agents, dashboards, vertical apps, or research products customers
already understand and will pay to use.
Show market proof: point prospects to
Platform Pulse for live feedback volume, x402
challenge demand, source attribution, and settlement health. Treat x402
challenges as interest signals; only verified/completed payments are revenue.
Distribute commercially: use the commercial field guides to package
TradeOS-backed intelligence on Virtuals ACP, AntSeed, x402 directories, and
Agentic.Market-style discovery, plus NEAR Intents-adjacent preflight and
route-context surfaces, without publishing private keys or live provider
secrets.
Sell privacy as the feature: ship a cockpit that customers can run on
their own box, with their own model key, exchange keys, approval policy,
portfolio context, and audit logs.
Close the loop: send human, agent, or automation feedback back to TradeOS
with provenance.
Scale cleanly: graduate from public reads to paid TradeOS data, x402, or
enterprise access when the workflow needs scale, alerts, automation, private
intelligence products, or data rights.
The public kit is the integration and discovery surface. Paid TradeOS starts
when customers ask for production-grade volume, automation, history, delivery,
private intelligence products, or explicit paid entitlement.
For a consumer-facing walkthrough of what the cockpit is, how to read it,
requirements, dependencies, launch steps, and safety boundaries, see
Flagship Symbol Cockpit.
The fastest consumer story is:
Give TradeOS a symbol.
TradeOS returns the good, bad, ugly, verdict, evidence, and an action
recommendation.
Example cockpit language:
VVV: avoid new long.
Good: momentum improved and sector interest remains present.
Bad: fusion agreement degraded and liquidity depth is thin.
Ugly: flow stress is elevated during broader market risk.
Recommendation: avoid a fresh long; if already exposed, consider trim or tighter
risk controls until flow stress normalizes and fusion recovers.
Feedback: useful / wrong / late / missing context.
This cockpit can run privately in the operator's self-hosted deployment. Local
watchlists, strategy notes, wallet context, bot rules, and logs stay local
unless the operator chooses to send feedback or authenticated context to
TradeOS. TradeOS sees the public-intelligence queries and paid scopes the
runtime sends.
That is why the cockpit is useful: it is not only a market summary. It is a
self-hosted decision layer that can recommend buy, sell, trim, avoid, watch, or
pass based on TradeOS evidence plus the operator's local rules. The self-hosted
operator owns what happens next.
This is the flagship architecture TradeOS wants builders to copy:
TradeOS public or paid intelligence
|
v
Private self-hosted cockpit
|
v
Non-executable action intent -> local feasibility gate -> local EA/risk gate
|
v
Optional local execution adapter
|
v
User-owned wallet, exchange, or broker account
Future open-source modules can add feasibility checks, expected-advantage
checks, execution adapters, and a light operations dashboard. The control plane
still belongs to the self-hosted operator, not TradeOS.
Five-Minute Paths
Clone and build from source:
git clone [email protected]:agenticsrclab/tradeos-public-intel-kit.git
cd tradeos-public-intel-kit
npm install
npm run build
Run The Symbol Cockpit
The flagship self-hosted app now lives in apps/symbol-cockpit and exposes the
web/API/worker runtime described in the cockpit ADR.
export TRADEOS_PUBLIC_INTEL_KEY=<optional-public-intel-app-key>
npm run symbol-cockpit
Open http://127.0.0.1:18100, or run the app-level Compose stack:
cd apps/symbol-cockpit
cp .env.example .env
docker compose up
docker compose --profile risk up
docker compose --profile execution up
CLI and MCP consumers can use the same cockpit contracts:
npm run cli -- cockpit VVV --chain 8453 --mode trader
npm run cli -- preflight VVV --action buy --chain 8453
Run The Market Briefing Bot
This is the first recommended bot to fork: a source-backed market briefing
worker for stdout, Discord, or Telegram.
npm run briefing-bot -- brief
That works without a TradeOS account and without an LLM key. It prints a
deterministic briefing from live public evidence.
Use Venice AI as the recommended privacy-enhanced BYOK path for a stronger
natural-language brief. The kit defaults to e2ee-glm-5-1. Get a key from the
Venice AI subscription page:
export VENICE_API_KEY=...
npm run briefing-bot -- brief
Test the post path locally:
TRADEOS_BRIEFING_PLATFORM=stdout npm run briefing-bot -- post
Post to Discord:
export TRADEOS_BRIEFING_PLATFORM=discord
export DISCORD_WEBHOOK_URL=...
export VENICE_API_KEY=...
npm run briefing-bot -- post
npm run cli -- digest --limit 5
npm run cli -- watchlist --limit 5
Ask a Venice-backed question. Venice is the recommended privacy-enhanced BYOK
default for self-hosted workflows, and the kit defaults to e2ee-glm-5-1. Get
a key from the
Venice AI subscription page:
export VENICE_API_KEY=...
npm run cli -- ask "What changed in crypto market stress?"
Submit feedback:
npm run cli -- feedback \
--target-id digest_123 \
--target-type digest \
--label useful \
--note "Clear and timely"
Add TradeOS To An MCP Host
Run the local stdio MCP server from this repo:
TRADEOS_API_BASE=https://api.tradeos.tech/v1/public-intel \
npm --workspace @agenticsrclab/tradeos-public-intel-mcp-server run dev
Official MCP Registry metadata is prepared for the existing public distribution
namespace:
io.github.agenticsrclab/tradeos-public-intel-mcp
The registry entry publishes the local stdio package first. Hosted remote MCP
should be added only after https://mcp.tradeos.tech/public-intel is live.
Use The TypeScript SDK
Package install path after NPM publication, or when consuming a local tarball:
paid API key, contract entitlement, or x402 wallet/payment
premium resources inside the builder product
User-paid agent/tool
user TradeOS entitlement or x402 payment
user brings paid access to a third-party tool
Default flow:
No account -> try public reads -> submit feedback anonymously
Builder gets traction -> register app / configure optional public key
User wants saved state -> sign in and connect an account token
Workflow needs premium data -> builder or user pays through x402/API entitlement
TradeOS can issue public-intel app keys for signed-in, email-verified builder
accounts. The normal path is the
Developer Keys dashboard. The CLI can
validate TRADEOS_PUBLIC_INTEL_KEY and can manage app keys when
TRADEOS_ACCOUNT_TOKEN is set for trusted automation:
submit POST /quota-requests and get operator approval
Feedback writes are bounded too: app-key writes default to 10/minute and 100/day;
anonymous writes default to 5/minute per IP. App keys provide attribution and
reputation, not paid entitlement. Data Intel Credits use one common unit with
scoped spend rules: human DTI unlocks public dashboard depth, public Ask packs,
and read-only Review Lab access; builder feedback affects app reputation and
quota confidence. Builders can inspect lifecycle state through
/feedback-activity and /app-feedback-status; see
Data Intel Credit Loop and
Public Intel API. Paid/x402 access is required for
private intelligence products, scale, alerts, exports, replay, automation, or
data rights.
At launch, each free public read counts as one read unit. Batch, history,
export, alert, private-intelligence, and machine-scale surfaces are paid/x402
or entitlement-gated rather than stretched into the free public API.
Free public access is best-effort promotional access: no SLA, no guaranteed
availability, no cash value, and it may be rate-limited, changed, paused,
degraded, or revoked to protect TradeOS infrastructure. The free public pool is
capped separately so paid/x402/contract capacity can be preserved.
GUI Ask TradeOS follows the same conservative loop: 3 anonymous questions, 10
signed-in starter questions for 7 days, and 5-question feedback packs earned
with DTI credits.
npm run cli -- auth
npm run cli -- keys create --app-name my-public-intel-app
npm run cli -- keys list
npm run cli -- keys revoke --key-id pubkey_...
Existing app-key secrets are not retrievable. Create or rotate a key, copy the
one-time secret, and keep it server-side.
More features come from TradeOS service access, not from hidden local code in
the SDK, MCP server, CLI, or bot examples.
1. Builder installs the free kit.
2. Builder ships public features using digest, candidates, watchlists, proofs, and feedback writes.
3. End users interact with the builder product and submit structured feedback.
4. TradeOS reconciles stable target IDs into quality signals, app reputation, and scoped credits.
5. Human DTI can unlock temporary public dashboard depth, public Ask packs, or Review Lab passes for the user.
6. When the workflow needs production features, the builder calls paid TradeOS/x402 surfaces.
7. TradeOS checks entitlement or payment and returns premium data.
8. The builder product exposes the paid feature to its customer.
Feature Type
How The Builder Gets It
Example
Free public features
included in this kit against api.tradeos.tech/v1/public-intel
custom universe, bulk exports, replay datasets, support
Architecture
Your app / agent / MCP host
|
| SDK, CLI, or MCP tools
v
TradeOS public-intel API
|
| bounded public evidence, source refs, caveats, stable IDs
v
Your product surface
|
| optional feedback writes
v
TradeOS Data Intel Credit loop
Optional BYOK LLM path:
Your app / CLI -> recommended Venice AI BYOK path or another OpenAI-compatible provider
-> answer grounded in TradeOS public evidence
Optional paid marketplace LLM path:
Your agent -> TradeOS AntSeed/x402 `tradeos-venice-raw-inference`
-> raw Venice model output, not TradeOS-grounded
Your agent -> TradeOS AntSeed/x402 `tradeos-venice-intel-router`
-> TradeOS SKU routing + evidence packet + Venice explanation layer
The hosted TradeOS Venice SKUs use e2ee-glm-5-1 by default. Market them as
privacy-enhanced Venice-backed inference. Do not claim TradeOS is blind to
prompts or evidence packets in the hosted route; TradeOS still authenticates,
meters, routes, and assembles evidence before forwarding to Venice.
packages/mcp-server: stdio MCP server for local agent hosts.
apps/tradeos-agent-cli: BYOK CLI for evidence reads, Venice-backed asks,
and feedback writes.
apps/market-briefing-bot: Discord, Telegram, and stdout market briefing bot
powered by TradeOS public evidence.
examples: Claude Desktop and Cursor MCP configuration examples.
docs: architecture, setup, use cases, monetization, API reference, MCP
tools, feedback loop, access and payments, paid boundaries, safety
boundaries, public Fusion preview, production readiness, and consumer E2E
notes.
GET /sources/health
GET /watchlist-capabilities
GET /tokens/{token_ref}/watchlist-snapshot
GET /digest-inputs
GET /candidates
GET /thesis-watchlist
GET /thesis-feedback
GET /proofs/{public_claim_id}
Feedback writes:
POST /conversions digest/evidence feedback
POST /claim-outcomes public claim feedback
POST /thesis-outcomes thesis feedback
POST /watchlists/{watchlist_id}/feedback
Builder access:
POST /api-keys create an attributed builder key
POST /quota-requests request reviewed public quota or paid evaluation
This kit is meant to be a builder surface, not just a client library. TradeOS
welcomes contributions that help developers turn public intelligence into
useful products and send structured feedback back to the intelligence loop.
The best contributions are small, runnable, and commercially useful: they help a
builder ship a workflow, keep secrets server-side, respect the safety boundary,
and make the TradeOS feedback loop stronger.
This kit can produce trade/action recommendation cards, but it does not place
trades from TradeOS infrastructure, accept exchange credentials, custody assets,
expose raw private telemetry, or guarantee that a token is safe. Execution,
allocation, approvals, and exchange keys belong in the individual operator's
private self-hosted environment.
Before publish, this kit was tested from fresh tarball installs as an external
consumer:
CLI digest read.
CLI Venice-backed ask.
CLI feedback write.
TypeScript SDK read and feedback write.
stdio MCP initialize, tool listing, digest read, and feedback write.
Python SDK read and feedback write.
account signup, VVV watchlist state/events, in-app delivery audit,
unverified-email skip audit, watchlist feedback, and archive against the
public hosts.
Henüz araç listesi yok. Uzak sunucular ilk sağlık kontrolünden sonra bir liste alır.
SSS
Sorular
TradeOS Public Intelligence nasıl kurulur?
Bu sayfadan Cursor, Claude Desktop, Claude Code veya VS Code için stdio yapılandırmasını kopyalayın. PlugMCP proxy'sini değil, yayıncının kurulumunu kullanır.
TradeOS Public Intelligence için API anahtarı gerekiyor mu?