Decision infrastructurep50 141 ms · 99.99% · SF / NY

The decision layer for engineering organizations.

Approvals, prioritization calls, and escalations over a single REST API — observable, idempotent, versioned. Integrate once; every decision after that is a request.

Backed by Horowitz Andreessen and Eventual Ventures

$ curl https://api.mngr.dev/v1/decisions \
    -u sk_live_51NqRk2…: \
    -d context="Reassign two seats to Growth?"

{
  "id": "dec_4g8Xk2",
  "object": "decision",
  "decision": "approve",
  "rationale": "Consistent with current headcount plan.",
  "confidence": 0.74,
  "latency_ms": 141
}
Decisions in production at
  • Meridian Aggregate Holdings
  • Nordvik Logistics AS
  • Amalgamated Substrate Corporation

01 · The decision layer

Your organization already runs on decisions. It just runs them on people.

Every engineering org operates a decision layer: approvals, prioritization, escalation, performance conversations. Today that layer is implemented in humans. It has no SLA, no observability, and a load profile that has been getting worse for seventy years.

mngr moves it to infrastructure. Managerial decisions — approvals, prioritization calls, one-on-ones, performance conversations — are served over REST, with the delivery guarantees the rest of your stack already has.

Management-layer load, selected figures
Average CEO span of control, mid-1980s → mid-2000s25 → 10
Managers reporting burnout “very often” or “always”, 2021635%
Favorable rulings, start → end of one unbroken decision session365% → ≈0%
Correlation of administrative growth with workload10.00 · stable since 1955

02 · Platform

Management, with the properties of infrastructure.

  • Decisions API

    POST context, receive a decision. Every response carries a rationale, a confidence score, and a 90-day reversal window.

  • Approval routing

    Chains resolve server-side, in order, with per-hop timeouts. Idempotent by default: submitting the same request twice cannot produce two different answers.

  • Decision observability

    Trace any decision from request to rationale, with latency percentiles per decision class.

  • Escalation queues

    Requests outside the standing decision envelope escalate automatically. Tier-2 turnaround is covered by SLA.

03 · Console

Try the API.

The sandbox runs the same decision engine as production, pinned to API version 2026-06-15. No signup required.

mngr consolesandbox · api version 2026-06-15
# POST https://api.mngr.dev/v1/decisions
# Authorization: Bearer sk_test_51NqRk2… (sandbox)
#
# Sandbox decisions are non-binding.
# Send the request to receive a response.

04 · In production

“We no longer wait to find out what we think.”

VP Operations Infrastructure, Meridian Aggregate Holdings

Meridian Aggregate Holdings routes 4,100 decisions a day through mngr, consolidating the approval surface of operations in nineteen countries onto a single decision plane. Median decision latency fell from 11 days to 210 ms.

05 · Pricing

Metered per decision-unit.

One decision-unit (DU) corresponds to one standard managerial determination. Sandbox is free. Team is $749/month and includes 25,000 DU. Enterprise adds dedicated manager instances, SCIM, and a 99.99% SLA.

Volume pricing decisions are made through the API.

Consumption schedule, API version 2026-06-15
Standard approval1 DU
Prioritization call3 DU
One-on-one (30 min)12 DU
Performance conversation40 DU
Decision reversaloriginal cost + 1 DU

06 · Reliability

Decisions are infrastructure. We publish our numbers.

Service levels, trailing 90 days
Surfacep50p99SLA
Decisions141 ms340 ms99.99%
Approvals96 ms220 ms99.99%
Webhooks1.2 s4.0 s99.9%
Escalations (tier 2)4 business days9 business days99.9%

Same context, same decision: requests are idempotent, and a replayed question returns the original response with idempotency-status: replayed. Certified ISO 9001:2031 and SOC 2 Type III.

07 · Access

Get an API key.

Sandbox keys are instant. Production access is metered and reviewed: leave a work email and a decision will be made about your account.