All systems operational--:--:-- UTC

Five modules. One data model.

AI, Robotics, Payments, Analytics and Trading share one event store, one permission model and one audit log. Use one module or all five; they are built to work together from the first event.

SOC 2 Type II and ISO 27001 certified. PCI DSS Level 1 for Payments. EU and US data residency.

AI

Agents that plan, call tools and wait for a human when the policy says so.

Define an agent as a model, a set of tools and a policy. Every run is recorded step by step: what it read, what it decided and what it changed, with latency and token cost on each step. Actions above a threshold you set pause at an approval gate instead of executing.

Typed tools
Tools are JSON Schema functions over your own APIs and every other ConsciousLab module. Arguments are validated before a call leaves the sandbox.
Approval gates
Per-tool policies by amount, scope or confidence. Approvers get the full trace and decide in the app, in Slack or by API.
Replay and evals
Re-run any recorded trace against a new prompt or model and diff the decisions before you promote it.
Streaming
Server-sent events for tokens, tool calls and state changes, so operators watch a run as it happens.
Run run_7Hq2xV, support-triage agent Awaiting approval
Model
cl-reason-2
Elapsed
1.44s
Tokens
2,184
  1. Sense
    tickets.search
    status:open priority:high, 3 results
    142ms
  2. Sense
    orders.lookup
    ORD-88214, Halvorsen Logistics
    88ms
  3. Reason
    plan
    Damaged pallet confirmed by scan RX-4471. Refund within policy (< €250).
    1210ms
  4. Act
    refund.issue
    €182.40 to card ending 4410
    gate
Policy refunds.max_eur = 250 requires one approverRejectApprove

Specifications

Models
cl-reason-2, cl-fast-1, or bring your own via OpenAI-compatible endpoint
Context
Up to 400k tokens per run, with retrieval over workspace data
Tool timeout
30 s default, 300 s maximum, retries with backoff
Trace retention
30 days on Pro, configurable to 7 years on Enterprise
Data residency
EU (Frankfurt) or US (Oregon); prompts are never used for training

Robotics

Fleet telemetry, remote control and staged firmware for every unit in the field.

Connect AMRs, arms, drones and fixed sensors through a ROS 2 bridge or the lightweight edge agent. Telemetry streams at up to 100 Hz per channel into the same store the rest of the platform reads, so an agent can see a motor running hot and open a maintenance ticket in the same run.

ROS 2 bridge
Subscribe to topics and expose services without rewriting nodes. Humble and Jazzy are supported; DDS traffic stays on your network.
Thresholds and alerts
Per-channel limits with hysteresis. Breaches become events that Analytics charts and agents can act on.
Staged firmware
Roll out to 1%, 10%, then the fleet, with automatic halt if fault rate rises above baseline. Images are signed with Ed25519.
Remote control
Pause, resume, return to dock and e-stop from the console, each action authorised per role and written to the audit log.
Fleet, Rotterdam + Leipzig + Fremont2,418 units online
UnitSiteStateBatteryMotor °CFirmware
AMR-0412Rotterdam DC2Moving71%41.24.12.3
AMR-0415Rotterdam DC2Charging28%36.84.12.3
ARM-1180Leipzig Cell 4Picking100%47.92.8.0
AMR-0398Rotterdam DC2Fault54%62.44.11.9
QDR-0031Fremont YardIdle88%30.11.4.2
AMR-0398 drive motor temperature, last 6 hThreshold 60.0 °C

Specifications

Protocols
ROS 2 (DDS), MQTT 5, gRPC, OPC UA (read)
Telemetry
Up to 100 Hz per channel, 13 months hot retention
Edge agent
arm64 and x86_64 Linux, 38 MB, buffers 72 h offline
Command latency
p95 180 ms from console to unit on LTE
Safety
E-stop is never gated by policy; supports ISO 10218 cell layouts

Payments

Multi-currency wallets, invoices and settlement on a double-entry ledger.

Every movement of money is two ledger lines that must balance. Invoices, refunds, FX conversions and payouts all post to the same ledger, which means finance can reconcile to the cent and an agent can never issue a refund that the books cannot explain.

Wallets
Hold balances in 34 currencies with available and pending amounts tracked separately. FX at mid-market plus a disclosed spread.
Invoices
Draft, send, remind and reconcile. Line items, tax rates and payment terms per customer; PDF and e-invoice (XRechnung, Peppol) output.
Payouts and refunds
SEPA Instant, SEPA Credit Transfer, ACH and card refunds, with idempotency keys on every call.
Webhooks
Signed with HMAC-SHA256, retried for 72 hours with exponential backoff, and replayable from the console.
Ledger, Northwind Robotics GmbHBalanced, 0.00 drift
EUR wallet
1,284,902.17
avail 1,201,440.00
USD wallet
846,210.55
avail 846,210.55
GBP wallet
92,417.80
avail 88,017.80
UTCReferenceAccountDebitCreditStatus
14:02:11
INV-2049
Halvorsen Logistics
receivables18,400.00Settled
14:02:11
INV-2049
Halvorsen Logistics
cash.eur18,400.00Settled
13:47:52
RFD-0193
Refund, ORD-88214
cash.eur182.40Pending
13:21:09
FX-55120
EUR to USD at 1.0874
fx.clearing40,000.00Settled
12:58:30
PO-7781
Kessler Antriebstechnik
payables6,912.00Scheduled

Specifications

Compliance
PCI DSS Level 1 service provider; card data never touches your servers
Rails
SEPA, SEPA Instant, ACH, Faster Payments, cards via licensed acquirers
Currencies
34 for wallets, 120+ for card acceptance
Ledger
Immutable, double-entry, exportable to DATEV, NetSuite and CSV
Settlement
T+0 for SEPA Instant, T+1 for cards in the EU and US

Analytics

Metrics, funnels and cohorts over every event the platform produces.

Agent steps, telemetry samples, ledger postings and order fills are all events in one columnar store. Ask questions that cross module boundaries, such as how approval latency affects refund cost, without exporting anything to a warehouse first.

One event model
Every module writes typed events with a shared entity, actor and timestamp. Your own events use the same schema.
Funnels and cohorts
Build them in the editor or in SQL. Results are cached and recomputed incrementally as new events land.
Dashboards
Shareable, versioned, and refreshable on a schedule. Every tile links back to the query that produced it.
Export
Continuous export to S3, GCS, BigQuery and Snowflake in Parquet, partitioned by day and module.
Approval funnel, last 24 hEvents ingested 3.92B
  1. Agent run started184,220 100.0%
  2. Tool call succeeded171,032 92.8%
  3. Approval requested23,418 13.7%
  4. Approved within SLA21,904 93.5%
  5. Action settled21,611 98.7%
Workspace retention by signup week
WeekW0W1W2W3W4W5
Aug 04
100
71
63
58
55
53
Aug 11
100
74
66
61
57
Aug 18
100
69
62
59
Aug 25
100
76
68
Sep 01
100
78
Ingest, events/s
45,370

Specifications

Ingest
Sustained 50k events/s per workspace, burst to 250k
Query
p50 340 ms over 30 days of data for typical funnels
SQL
ANSI SQL with window functions, read-only, row-level security
Retention
13 months on Pro, unlimited on Enterprise
Privacy
Field-level PII tagging, deletion requests honoured within 30 days

Trading

A professional terminal with pre-trade risk checks on every order.

Real-time equities, ETFs, FX and crypto with depth, charts and an order ticket that supports brackets and OCO. Every order passes buying power, position, loss and price-band checks before routing. Paper trading runs the same engine against live quotes.

Market data
Level 1 and Level 2 for US equities, consolidated crypto books from six venues, and FX from tier-one liquidity providers.
Order types
Market, limit, stop, stop-limit, trailing stop, bracket and OCO, with day, GTC, IOC and FOK time in force.
AI signals
Model-generated signals with the features behind them and their historical hit rate. Informational only, never auto-executed by default.
FIX and REST
FIX 4.4 sessions for institutional flow, REST and WebSocket for everyone else.
NVDA, Nasdaq 189.42 +1.84%
PriceSize
  • 189.461,200
  • 189.45820
  • 189.442,400
  • 189.43310
  • 189.421,640
Spread 0.01 (0.5 bp)
  • 189.41900
  • 189.402,100
  • 189.39640
  • 189.381,780
  • 189.373,020
BuySell
Type
Limit, day
Limit
189.40
Quantity
250
Notional
$47,350.00
Bracket
TP 196.00 / SL 184.20
Pre-trade checks
  • Buying power$212,480 avail
  • Position limit4.1% of book
  • Daily loss limit$3,120 of $10,000
  • Fat-finger bandwithin 2.0%

Specifications

Coverage
11,400 US equities and ETFs, 62 FX pairs, 140 crypto pairs
Latency
Median 4.1 ms order acknowledgement from our New Jersey edge
Risk
Pre-trade checks on 100% of orders, kill switch per account
Protocols
FIX 4.4, REST, WebSocket
Paper
Unlimited paper accounts, reset at any time

Sense, reason, act, on one record of what happened.

The modules do not sync with each other; they write to and read from the same event log. That is why a motor temperature reading, the agent decision it triggered and the work order it paid for can be traced in one query.

One event, as every module sees it
{
  "id": "evt_01J9ZK4R7W2Q",
  "type": "telemetry.threshold_breached",
  "module": "robotics",
  "entity": "unit/AMR-0398",
  "actor": "system/edge-agent",
  "at": "2026-10-17T13:47:52.118Z",
  "data": { "channel": "drive.motor_temp_c", "value": 62.4, "limit": 60.0 },
  "links": ["site/rotterdam-dc2", "workorder/WO-3318"]
}
  1. Sense

    1/3

    Modules emit typed events: a telemetry sample, a ledger posting, a fill, a tool result. Each carries an entity, an actor and a timestamp.

    • telemetry.sample
    • ledger.posted
    • order.filled
  2. Reason

    2/3

    Agents and rules read the same events and the current state of each entity, then propose an action with the evidence attached.

    • run.step
    • signal.generated
    • policy.evaluated
  3. Act

    3/3

    Actions go through the owning module, its permissions and its policy gates. The result is a new event, so the loop closes and is measured.

    • refund.issued
    • unit.paused
    • order.routed
Identity
One workspace, SSO via SAML 2.0 or OIDC, SCIM provisioning
Permissions
Role-based with per-module scopes; API keys carry the same scopes
Audit
Every action, human or agent, in one append-only log with actor and IP
Encryption
TLS 1.3 in transit, AES-256 at rest, customer-managed keys on Enterprise
Availability
99.95% monthly uptime SLA on Enterprise, 14 regions

Start with one module. Add the rest when you need them.

Starter is free for up to three seats and 1 million events a month.