v2.0 — Agentic evals are live
Open the black box of AI.
BLACKBOX07 routes every prompt to the right model, verifies every answer and shows you exactly why — through one API your team can actually trust.
SOC 2 Type II
GDPR ready
99.99% uptime
Plugs into 40+ models and the stack you already run
Capabilities
Six systems. One black box.
01
Neural Routing
Every request finds the fastest model path in real time.
Hover to decrypt
12ms
Median routing latency
A learned router scores 40+ models per call and picks the cheapest one that clears your quality bar.
02
Guardrails
Policy checks run on every token before it leaves the box.
Hover to decrypt
0
Unreviewed outputs in prod
PII redaction, jailbreak detection and custom policies — enforced inline and logged forever.
03
Live Evals
Score quality continuously against your own golden sets.
Hover to decrypt
1.2M
Evals run per day
Ship prompts like code: every change is scored on regressions, cost and latency before rollout.
04
Model Mesh
Frontier and open models behind a single interface.
Hover to decrypt
40+
Models, one endpoint
Swap providers without a redeploy. Fallbacks kick in automatically the moment a model degrades.
05
Full Traces
See every hop, token and decision the box made.
Hover to decrypt
100%
Requests traced
Replay any request, diff two runs and export traces to the tools your team already uses.
06
Edge Runtime
Run agents close to your users, everywhere.
Hover to decrypt
32
Global regions
Cold starts under 50ms, autoscaling to zero and data residency controls per workspace.
How it works
Unbox production AI in four moves.
Scroll to watch each layer unfold. Most teams go from first call to fully evaluated traffic in a single afternoon.
Average setup: 37 minutes
01
Connect
Drop in one SDK and point your existing OpenAI‑compatible calls at the box. No rewrites.
$
npm i @blackbox07/sdk
02
Define
Describe quality in plain language or upload golden sets. The box turns them into live evals.
$
evals: [faithful, on-brand, p95<800ms]
03
Route
Every call is scored, routed and guarded in under 15ms — across 40+ models.
$
route: auto // cheapest model that passes
04
Observe
Traces, costs and regressions stream to your dashboard the second they happen.
$
traces → datadog · grafana · slack
Use cases
Built for teams that ship on Friday.
Move your cursor across the cards — they respond. So does everything you build on BLACKBOX07.
Where’s my order #4821?
It left our Berlin hub 2h ago and arrives tomorrow before 10:00. Want live tracking?
Support
Copilots that actually resolve
Deflect 68% of tickets with answers grounded in your docs, orders and policies — escalating the rest with full context.
agent.plan(ticket)
→ read repo/src/billing
→ patch 3 files, +42 −17
→ test 128 passed
Engineering
Code agents with guardrails
Plan, patch and test inside a sandbox. Nothing merges without passing your evals.
98.7%
Field accuracy on 2M invoices
Documents
Extraction you can audit
Every field comes with a confidence score and a trace back to the source pixel.
Realtime
Voice agents under 300ms
Streaming speech-to-speech with interruptions, tool calls and live transcripts — routed to the fastest model per turn.
Developer experience
Ten lines. Zero lock‑in.
OpenAI‑compatible by default, typed end to end, and portable across every provider. Your prompts stay yours.
SDKs for TypeScript, Python, Go and Rust
Streaming, tools and structured outputs
Bring your own keys — or use ours
route.ts
TS
01
import { Box } from ″@blackbox07/sdk″
02
03
const box = new Box({ key: env.BB07_KEY })
04
05
const res = await box.run({
06
prompt: ticket.body,
07
route: ″auto″,
08
evals: [″faithful″, ″on-brand″],
09
guard: { pii: ″redact″ }
10
})
200 OK · 142ms · routed → claude · evals 2/2 passed
40+
Models behind one API
12ms
Median routing overhead
99.99%
Uptime, trailing 12 months
3.2B
Requests routed every month
Wall of love
2,400 teams stopped guessing.
“We replaced four vendors and a homegrown eval stack with BLACKBOX07. Latency dropped by half — and our infra bill followed.”
Maya Okafor
CTO, Loopwise
“The traces alone are worth it. For the first time our PMs can see why the model answered the way it did.”
Jonas Lindqvist
Head of ML, Fernhill Labs
“Routing moved 70% of our traffic to cheaper models without a single quality regression. That’s real money.”
Arjun Mehta
VP Engineering, Quanta Freight
“Guardrails that run inline, not as an afterthought. Our compliance team signed off in a week.”
Lena Park
Staff Engineer, Helix Health
“We ship prompt changes like code now: evals on every PR, instant rollback. It changed how we work.”
Tomás Rivera
Founder, Parcelio
“Voice agents under 300ms in 12 regions. Our customers think they’re talking to a human.”
Ines Laurent
Product Lead, Stackmint
Pricing
Pay for answers, not for guesses.
Start free, scale when it works. Every plan includes routing, traces and the full SDK.
Monthly
Yearly
−20%
Starter
For side projects and prototypes.
$0
/ month
free forever
10k requests / month
5 models
Basic traces, 7 days
Community support
Pro
Most popular
For teams shipping AI to real customers.
$49
/ month
billed monthly
1M requests / month
All 40+ models
Live evals & guardrails
Traces, 30 days
Priority email support
Scale
For high-volume and regulated workloads.
$199
/ month
billed monthly
Unlimited requests
Dedicated routing capacity
SSO, SCIM & audit logs
EU / US data residency
99.99% uptime SLA
FAQ
Questions from inside the box.
What exactly is inside the black box?
Which models can I route to?
Do you store my prompts or outputs?
How does routing decide which model to use?
Can I self-host BLACKBOX07?
What happens if a provider goes down?
Ready to open the box?
Route your first million tokens free. No credit card, no sales call, no black magic.