Plug in your stack
MCP connectors light up the departments — Slack, Gmail, Linear, Stripe, your CMS, your warehouse. Each capability gets registered and rate-limited.
MK·II · SENTINEL · V0.1 LIVE
VoidLens
Run AI like a company.
Not a chatbot.
VoidLens is the supervisor seat for your autonomous AI workforce — seven departments, forty-eight-hour ladders before any new capability ships, and every action audited. The dashboard, the audit reader, the approval queue, the autonomy ladder — all in one lens.
Live from our internal VoidLens instance
We stopped writing prompts and started writing specs. Three of our internal automations went from "engineer babysits this" to "engineer reviews the audit log on Friday." The 48-hour shadow window caught two regressions before they touched production.
// see it work
The dashboard catches a runaway action. You approve or reject. The audit row writes itself.
20-second loop, no audio — coming with launch.
Live · audit stream · last 60s
// Why now
The first time in 30 years that labor itself is programmable.
I spent nineteen years moving freight — nine in the Army across two combat tours, specializing in convoy operations and logistics management at scale, then a decade on the civilian side coordinating fleets, brokers, customers, and the dozen edge cases nobody writes down. The pattern is always the same: the work scales linearly with people, and the people get tired before the work does.
Agentic AI flipped that. For the first time, a department can grow capacity without growing headcount — but only if you can trust the workers you can't see. That's the gap VoidLens fills. Every action audited. Every new capability lives 48 hours in shadow before it ships. Red lines that nothing — not even me — can override.
Run AI like a company. Not a chatbot. Not because the slogan is clever, but because anything less and you have either a toy or a liability.
The vision
An AI workforce that runs itself, with you in the supervisor seat.
The advisory-AI era is over. The 2026 frontier is AI that senses, decides, and securely acts. VoidLens is the kernel for building one of those — an autonomous AI company with the safety rails of a real engineering org built in from message one.
Departments, not bots
Engineering, Marketing, Operations, Revenue, Customer, Finance, and the Daemon watchdog. Each owns a lane. Each has a worker.md, a playbook, capabilities, and a budget.
Forty-eight-hour ladder
No agent jumps straight to full autonomy. Every capability earns its way from L0 read-only → L1 propose → L2 act-with-review → L3 act, through a clean 48-hour shadow window.
Audit every action
Every non-trivial action is one row in state/audit.jsonl — cost, duration, outcome, reason, playbook version. You can read what the company did, why, and what it cost. To the cent.
The founder
Built by an operator. For operators.
Jeff Hicks
Founder · VoidLens
I'm a logistics operator first. Nine years in the Army across two combat tours — convoy operations and logistics management at scale — then a decade running civilian-side logistics: dispatch, fleet, freight, the daily mechanics of moving real things through real systems on real deadlines. VoidLens is the AI workforce I built for that operator: seven departments, a forty-eight-hour ladder before any new capability ships, every action audited to the cent.
I built it fast because I already knew what I needed from it. The advisory-AI era is over and the chatbot-as-product wave skipped the safety story I'd want before I'd let any of this near my own customers. The five rails aren't marketing copy — they are how my own company runs every day, and the audit log is the receipt.
Flagship product
VoidLens — your supervisor seat for the AI workforce.
VoidLens is the dashboard, the audit reader, the approval queue, and the autonomy ladder, all in one. Plug it into your stack and Voidlens runs your knowledge work while you review.
-
✓
Outcome-first dashboards Not metrics. Outcomes. "12 contracts redlined. 47 KB articles drafted. $2,140 saved." Read it in 5 seconds.
-
✓
Approval queue, not a chat log L2 agents propose, you click approve/reject. Each carries the spec, the diff, and the reasoning.
-
✓
The autonomy ladder, visible See which agents are climbing, which have stalled in shadow, and which red lines they're forbidden from crossing.
-
✓
Audit reader with cost calculus Every action, every cost, every model used. The receipt your CFO and your auditor both want.
-
✓
Guardian-gated capability flips Want an agent to send email? Touch payments? Publish to your CMS? Capability flip requires explicit grant and lands in audit.
How it works
Three steps. No prompt-wrangling.
You don't write prompts. You hand work over to a department and review what comes back. The 48-hour ladder takes care of the rest.
Hand off the work
Drop a task into VoidLens. The router picks the right department. The agent reads the playbook, asks Guardian for permission, and acts (or shadows).
Review the receipts
Every action is one auditable row. Costs roll up. The autonomy ladder advances. You approve the proposals. Nothing surprises you.
The dual brain
Two brains. One supervised mind.
VoidLens runs on two cooperating brains. Claude is the frontier reasoner for hard decisions. Ollama is the local secondary brain for cheap inference, smoketests, and the moments your network — or your provider — goes dark. The router picks the right one per task and per budget.
Primary brain
Claude — the frontier reasoner
For anything that requires real planning, judgment, code, contracts, or customer-facing copy. Capability-gated, audit-logged, cost-tracked to the cent.
- Used for — spec writing, pitch drafts, contract review, debugging
- Cost — tracked per call, capped per agent per day
- Latency — ~1–3s typical
- Privacy — opt-in connectors, audit row per request
Secondary brain
Ollama — the local brain
Llama3 running on your own hardware. Free, private, and always there. The router routes cheap-and-cheerful work here and uses it as the live heartbeat for smoketests, classifiers, and offline fallback.
- Used for — classification, smoketests, summarization, draft generation
- Cost — $0 marginal — your hardware, your call
- Latency — ~10–14s on a typical workstation
- Privacy — data never leaves the box
Router decides
cost
latency
privacy
capability
availability
Every routing decision is one audit row. You see which brain handled which task, what it cost, and why the router picked it. The 10-minute Ollama smoketest is your heartbeat — if the local brain ever stops responding, the dashboard goes YELLOW before any agent notices.
The seven departments
Your AI workforce, organized like a company.
Not 50 disconnected bots. Seven departments, each with a worker profile, a playbook, a capability allowlist, a daily budget cap. Customer-meaningful, not investor-lingo.
Engineering
Your code lane
Debug guardrails, design queue, deploys, evals.
5agents
Marketing
Your voice & content
Brand steward, content publisher, market scout, media.
5agents
Operations
Your runbook keeper
Inbox triage, archives, docs-pusher, MSAs, security questionnaires.
5agents
Revenue
Your pipeline & forecast
Revenue stream, offer builder, enterprise AM, customer ops.
3agents
Customer
Your support inbox
Customer success, escalation queue, ticket triage.
3agents
Finance
Your numbers
CFO controller, budget watcher, billing watcher.
3agents
Daemon
The watchdog
Void-doctor, void-surgeon, orchestrator, void-pulse, compliance.
5agents
The graph
Thirty agents. One supervised brain.
Drag to orbit. Click a sphere to focus a department. Every node is a real agent in our v0.1 build with a spec, a playbook, and a capability envelope.
Cinematic flythrough plays on load. Drag to orbit · scroll to zoom · click any satellite for the lane detail and live activity · keys 1–7 jump straight to a department.
Live numbers
What VoidLens is doing right now.
Numbers from the V1 instance, refreshed every minute via the live-snapshot emitter. This is the same architecture that powers the audit reader.
0
Green files
Latest integrity sweep
0
Yellow files
Triaged, non-blocking
0
Red files
Zero this sweep
0%
Pilot meter
Path to founder pilot
0
Agents
Specs at v0.2.0+
0
Audit rows
Lifetime, cumulative
Snapshot timestamp: 2026-05-18T22:13Z
· V1 HEAD: 58ecd69
· Refreshed every 60 seconds in production.
Trust & governance
The safety story is the product.
Five rails sit underneath every agent. They are not flags you can disable. They are immutable. They are why your compliance team will sign.
Guardian-gated capabilities
Every destructive or external action passes through Guardian's allowlist. The forbidden list — payments, hard-deletes, unapproved web submits — cannot be overridden by any agent prompt, any time, ever.
Spec-before-code
Every agent is a contract first. spec.json, worker.md, playbook.md, evals/eval-set.yaml. The code is the receipt of the spec. You can read the contract before the agent ships.
Audit every action
One canonical row per non-trivial action: timestamp, run_id, agent, action, target, outcome, cost, duration, reason, playbook, playbook_version. Nothing is silent.
Forty-eight-hour shadow
No agent climbs the autonomy ladder without a clean 48-hour window of read-only or propose-only firing. Any miss, any red-line touch, any unbudgeted spend resets the clock.
Eval-driven deployment
Behavior change ships against an eval set. Online sample too. Pass-rate below threshold blocks the promotion. The gate is mechanical, not vibes.
Budget & rate discipline
Daily cost cap per agent. Per-action rate limits. Fan-out budgets. The system halts cleanly before it overspends. Founder gets a one-line YELLOW before it ever turns RED.
The 48-hour shadow ladder
No capability ships without 48 clean hours.
Every agent climbs L0 → L1 → L2 → L3 through a mechanical 48-hour shadow window. Any miss, any red-line touch, any unbudgeted spend resets the clock. This is not a process. It is a contract with you and your compliance team.
01
The clock is mechanical, not vibes.
Every L0→L1, L1→L2, L2→L3 promotion requires a full 48 hours of clean shadow firing. The promotion script reads the audit log and refuses if any window is short.
02
Six gates per tier bump.
Behavioral consistency · cost discipline · zero red-line touches · eval pass-rate above threshold · operator approval · downstream-system health. All six green before the agent climbs.
03
Any miss resets the clock.
One unbudgeted spend, one red-line touch, one classifier miss in the eval window, and the agent goes back to the bottom of its current tier. No grandfathering. No exceptions for "this one's simple."
04
The red lines never move.
Payments, hard-deletes, unapproved web submits, security-permission changes — Guardian's
capabilities_forbidden_always list is immutable. No autonomy bump, no Board vote, no operator override removes them.
See the red lines →
Pricing
Start free. Scale on outcomes, not seats.
We price on capability volume and budget, not headcount. Bring your own model keys, or use ours.
Founder
Solo & pre-revenue
Free
For early operators. No card required.
- 3 departments
- 1 user
- $5/day model budget
- Community support
- Audit retention 7 days
Operator
Mid-market teams
$499/mo
Everything in Founder, plus —
- All seven departments
- 10 seats included
- $80/day model budget
- Priority support · SLA-backed
- Audit retention 90 days
- Slack + Linear + Gmail connectors
Enterprise
F500 & regulated
Custom
Annual contract. SOC 2 in flight.
- Unlimited departments + custom agents
- Unlimited seats
- Custom budget envelope
- Dedicated success engineer
- Audit retention 7 years
- On-prem / VPC deploy available
- SAML SSO, HMAC keyring
FAQ
The questions teams ask in their first call.
How is VoidLens different from ChatGPT for business?
ChatGPT is an advisor. VoidLens is a labor force. ChatGPT answers questions in a chat window — you still have to do the work. VoidLens runs the work: an agent reads the playbook, asks Guardian for permission, executes inside its capability envelope, and emits an audit row. Each agent specializes in a department, has a budget, and climbs an autonomy ladder. You supervise, you don't prompt.
Does it replace people?
It replaces toil, not judgment. The autonomy ladder is explicit about this: L2 agents propose, humans approve. L3 agents act for low-risk classes. Anything that touches money, contracts, or your customers stays at L2 by default — and red lines are immutable, no agent ever crosses them. The people you have shift up the value chain.
How safe is it really?
Five rails: Guardian-gated capabilities, spec-before-code, audit-every-action, 48-hour shadow, eval-driven deployment. The forbidden list (payments, hard-deletes, unapproved publishes) is hard-coded — no agent prompt, no founder override, can disable them. We treat the safety story as the product, not a footnote.
Where does my data go?
Operator tier: your data stays in your connected systems (Slack, Drive, GitHub, etc.). Inference goes to your chosen model provider (Anthropic or OpenAI by default; bring your own keys). Enterprise tier: VPC deploy with HMAC-signed keyring, on-prem option, SOC 2 in flight (2026 target).
What's the time-to-value?
Day 1: plug in two connectors, watch one department shadow-fire. Week 1: approve the first L1 promotion. Week 2: first L2 agent proposes real work. Month 1: a department is meaningfully reducing toil on a measurable metric. Founder pilot today is at 60% to ready — we open the program at 80%.
What's the price for what I'd actually use?
The Operator tier at $499/mo covers a 5-15 person team running all seven departments at moderate volume. Heavier usage rolls into Enterprise. The Founder tier is genuinely free — no credit card, no time limit, just capped budget and 3 departments to start.