Market valuation · 2026-05-30 · internal

What is Nyx worth?

Sober, evidence-tied read on Nyx's market position, where it differentiates, where it loses, and what to charge. Written by a Nyx research worker (opus 4.8). Not a pitch.

Built for engineering teams that need autonomous agents under their own safety policy.

Headline assessment

Nyx is a genuinely differentiated piece of software with a thin but real moat: a hardcoded, test-proven safety floor plus signal-driven self-iteration that no shipping competitor matches at this level of safety wiring. The lead pitch for every enterprise conversation is the three-tier Push Gate whose Tier-1 floor cannot be overridden by any operator config (proven by floor-cannot-be-lifted.test.ts) -- this is the answer to "what stops it from going rogue." BYO-Claude-Code-subscription is a first-class feature: teams bring their own Claude seat, Nyx orchestrates it, no token markup, no vendor lock-in. Price it as a per-seat team tool (not a metered cloud agent), because Nyx sells software and safety rather than tokens. Recommended structure: Community free for adoption, Team at $39/user/month for engineering teams, Enterprise at custom for self-hosted deployments with audit log exports and SLA. The system now stands at roughly 31k lines (19k source + 12k tests) across 264 commits, with a token-free status pulse (GET /status/pulse), a self-improvement reflector (GET /reflection/digest), a Verifier runtime gate (in queue), and 5 worker-deadlock root-cause fixes all shipped as of 2026-05-30.

Where Nyx Pro sits against the field.

Pro tier monthly prices as of 2026-05-28. Devin shows its base subscription; the cloud autonomy model is fundamentally different. Nyx is the only entry that includes a non-overridable safety floor at the floor price.

PRO-TIER MONTHLY PRICE (USD) $0 $10 $20 $30 $40 $50 $60 $19 Nyx Pro recommended $20 Cursor Pro inference incl. $20 Devin base + $2.25 / ACU $10 Junie Pro BYO key $10 Copilot Pro 300 req $39 Copilot Pro+ 1500 req $20 Replit Core credits $20 Windsurf Pro quota
Nyx Pro $19 sits one dollar under Cursor Pro, the closest peer. The discount is intentional: Cursor bundles inference, Nyx makes you bring your own model.

Where Nyx wins on substance.

Six features that materially separate Nyx from the closest competitors. Check = full support, dot = partial, dash = absent. Most differentiation comes from safety + self-iteration + local-first; the rest is convergent.

Feature
Nyx
Devin
Cursor
OpenHands
Aider
Hardcoded non-overridable safety floor
-
-
-
-
Self-iteration on isolated branches
-
-
-
Local-first, bring-your-own-model
-
-
Per-role model tier routing (cost)
-
-
-
Always-on Critic + Watchdog singletons
-
-
-
Control Room + Discord + phone control
-
-
Nyx is the only entry that ships all six. The safety floor and self-iteration are the unique wedges; the rest are nice-to-haves.

Open core. Four tiers. Team as the primary commercial unit.

Nyx is local-first and BYO-Claude-Code-subscription -- the operator does not resell inference. That means near-zero cost of goods and no lock-in for buyers. Price like a per-seat team license (Cursor seat, JetBrains plan), not a metered agent. Community drives adoption; Team and Enterprise are where the revenue lives.

Community
$0
run it yourself
Teams evaluating, individual developers, solo founders.
  • Full local orchestrator
  • Push Gate (floor + blacklist)
  • Discord bot + dashboard
  • Critic + Watchdog
  • Self-iteration
  • Model routing
  • Cost guardrails
Pro
$19 / mo
$190 / year
Individual power-users who need sync and tunnel infrastructure.
  • Everything in Community
  • Multi-machine memory sync
  • More concurrent workers
  • Hosted named tunnel + custom domain
  • Priority updates
  • Email support
Enterprise
Custom
from ~$15-25k / yr
Regulated orgs that cannot ship code to a vendor cloud.
  • Everything in Team
  • SOC2 compliance path
  • Policy-as-code Push Gate
  • Security review of the floor
  • On-prem install support
  • Dedicated CS

Why BYO-subscription is a feature, not a constraint: Teams using Claude Max or Pro already pay Anthropic directly. Nyx orchestrates that subscription without touching tokens -- no per-ACU markup, no vendor dependency. This is the strongest cost argument against Devin or any metered agent: a team's existing Claude seat becomes the compute layer, and Nyx is the safety and orchestration layer on top.

Five angles, ranked.

  1. Team-grade agent platform (the primary bet). Engineering teams that need autonomous agents under their own safety policy, with BYO-Claude-Code-subscription as a cost advantage over Devin. Team tier is the primary commercial unit; Community drives top-of-funnel. Best fit for the local-first BYO-model reality.
  2. "Governed autonomy for Claude Code teams." Target eng teams already using Claude Max/Pro who want orchestration, oversight, and audit logs without Devin's per-ACU markup. Strongest acquisition wedge for Team tier.
  3. Enterprise self-hosted dev agent. Compete with Devin on security + control + custom deployment. Real money, long sales cycle. The hardcoded safety floor is the key differentiator in regulated environments.
  4. Niche: "the SRE agent for your production bot." Always-on on-call for teams running production systems that break at inconvenient hours. Small TAM, high intent, natural upsell to Team tier.
  5. Acquisition or reference impl. Optionality. Anthropic might want this as a reference Agent SDK orchestrator build. Not a plan, just a tail outcome.

What could kill this, ranked.

High · existential
Anthropic builds this
A hosted Claude Code Agent that bundles orchestration + safety gate would subsume Nyx's core. Because Nyx is built on Claude Code, platform absorption is the highest-probability existential risk.
High · structural
Local-first ceiling
Single-machine constraint blocks the cloud + team story. If multi-machine sync cannot ship cleanly, Pro and Team value propositions are hollow.
Medium · competitive
Open-source substitution
If OpenHands, Hermes, or Aider bolt on a safety gate + dashboard, Nyx's wedge narrows fast.
Medium · product
Unproven reliability, key-person risk
One developer, roughly 31k lines (19k source + 12k tests across 264 commits), multiple real feature demos (yaga.co.za, status pulse, self-improving reflector), self-iteration validated across dozens of overnight self-modifications. No evidence of robustness at scale. n equals 1.
Medium · platform
Model-provider dependency
Claude Code / Codex pricing, rate limits, and ToS are outside Nyx's control. A ToS change on automated or headless use could break the product overnight.
Low · positioning
Brand confusion with Atlas
Operator already owns Atlas (FBM Sniper cloud product). Two Greek-mythology brands in the same family of products may dilute either.

Five concrete next steps.

  1. Land 5 paying beta users at $19/month from the Claude Code power-user community (Discord, Reddit, X). Revenue beats any survey.
  2. Post a Show HN ("I built an autonomous multi-agent orchestrator on Claude Code in 12 hours") and measure signups + willingness-to-pay signal in comments.
  3. Run a Van Westendorp pricing survey on the target segment to find the price-sensitivity band before locking numbers.
  4. Ship multi-machine sync first and gate it behind Pro. Clearest free-to-paid wedge, directly de-risks the biggest weakness.
  5. Add opt-in telemetry to Community to learn which features are actually used before pricing them.

Where these numbers come from.

Competitor pricing gathered via a deep research pass on 2026-05-28 and cited inline in the source markdown at docs/valuation/nyx-market-value-2026-05-28.md. Some URLs come from secondary aggregators rather than vendor canonical pricing pages. Before publishing externally, confirm each figure against the vendor's own page, especially the Devin ACU rate, the Copilot 2026-06-01 usage-based transition, and the Windsurf quota change. No customer testimonials or adoption numbers were invented; where evidence is thin (self-iteration track record, willingness-to-pay) the doc says so.