NexusGPT Agent-Invisible

Name: Agent Native Offer Audit — NexusGPT
Item: NexusGPT
Rating: 36
Author: Agent Native Offers

AUDIE Score: 36/100 · Audited 2026-04-09 · Website: https://agent.nexus · Machine-readable: JSON

Pillar Scores

P1 Signal Architecture — 9/25

P2 Clarity Stack — 8/25

P3 Trust Envelope — 6/20

P4 Velocity Triggers — 6/10

P5 Gravity Design — 7/20

Executive Summary

NexusGPT scores 36/100 — Agent-Invisible — primarily because it has the right infrastructure for human developers but has not yet added the machine-readable layer that AI agents require to evaluate, integrate, and trust it autonomously. Its strongest pillar is Velocity Triggers (6/10): the free tier and self-serve API keys allow an agent to begin using the platform without a human gate. However, the absence of /llms.txt, schema.org markup, a published OpenAPI spec, and any audit trail means that AI agents cannot discover, evaluate, or verify NexusGPT programmatically. The top priority is a /llms.txt file — a low-effort, high-impact change that immediately signals agent-readiness to the AI ecosystem.

Strongest Signals

Friction-Free Entry (P4-A: 4/5) — Free tier with 4 agents + unlimited conversations, self-serve API keys, no sales gate. An AI agent can discover, sign up, and start using NexusGPT without human intervention on entry plans.
Certifications (P3-A: 3/5) — SOC 2, ISO 27001, ISO 42001, and GDPR certification provides third-party validated trust signals. Y Combinator backing adds credibility.
Integration Breadth (P5-A: 3/5) — 1,500+ tool integrations and multi-modal knowledge upload creates meaningful switching cost and data gravity once an agent is configured.

Critical Gaps

No llms.txt / Agent Identity Layer (P1-C: 1/5) — /llms.txt returns 404. There is no structured, machine-readable file that an AI agent could parse to understand NexusGPT's capabilities, pricing, or integration patterns autonomously.
No Audit Trail for Agent Systems (P3-C: 0/5) — Zero evidence of machine-accessible transaction logs. An AI agent orchestrating NexusGPT-powered workflows cannot programmatically verify what was executed, when, and at what cost.
No Substitution/Fallback Rules (P2-C: 0/5) — No guidance on downtime behavior, fallback agents, or service unavailability — a critical gap for agents that need reliability guarantees before committing to a workflow dependency.
No Schema.org Markup (P1-A: 1/5) — Offer, Product, and AggregateRating schema entirely absent. AI agents relying on structured data signals cannot evaluate or compare this offer programmatically.

Priority Actions

Publish /llms.txt — +4 pts · P1 · Effort: Low
Publish OpenAPI spec for the NexusGPT REST API — +2 pts · P1 · Effort: Low
Add schema.org/Offer markup to pricing page — +3 pts · P1 · Effort: Low
Create machine-accessible audit log endpoint — +4 pts · P3 · Effort: Med

All 20 Criteria

P1-A Structured Data — 1/5
No schema.org Product, Offer, or Organization markup detected on agent.nexus. The domain redirected 308 from gpt.nexus with no structured data evidence in scraped content or search results.

P1-B Machine-Readable Pricing — 2/5
Pricing tiers exist as HTML content (Free/$0, Hobby/$19/mo, Pro/$79/mo, Enterprise/custom) with agent and message limits stated per tier, but not tagged with schema.org/Offer or any machine-readable format.

P1-C llms.txt / Agent Layer — 1/5
/llms.txt returns 404. Documentation exists at docs.gpt.nexus but is written for human developers, not structured for LLM consumption. No agent card published.

P1-D API / MCP Availability — 3/5
REST API documented at docs.gpt.nexus with four resource types (Contracts, Agents, Chats, Messages). API key authentication. Users can import OpenAPI schemas to power agent tools. However, no published OpenAPI spec for the NexusGPT API itself, and no MCP server or agent card.

P1-E Discoverability (GEO) — 2/5
Appears in AI tool search results. Docs indexed. No evidence of AI retrieval-optimized content or structured content designed for LLM indexing beyond standard SEO.

P2-A Offer Completeness — 2/5
What (agent platform) and How Much (pricing tiers) are findable separately but not unified in one machine-parseable source. Enterprise tier requires sales contact.

P2-B Scope & Limits — 2/5
Agent count limits (4/16/30 per plan) and message limits (10K/35K per plan) are stated per tier. API rate limits not published. Storage or tool limits not explicitly stated.

P2-C Substitution Rules — 0/5
No guidance on service unavailability, downtime behavior, or fallback logic found on any scraped page. No fallback or substitution rules stated.

P2-D Conditional Logic — 2/5
Enterprise plan requires contact sales. Some conditions like agent count caps are disclosed per tier. But conditions are scattered across pricing page prose, not machine-readable.

P2-E Semantic Precision — 2/5
Documentation uses some precise terms (API key auth, agent contract model, similarity threshold 0–1 range, retrieval limit 1–5 snippets) alongside vague marketing language ("AI that finally delivers"). Mix.

P3-A Verifiable Performance — 3/5
NexusGPT holds SOC 2, ISO 27001, ISO 42001, and GDPR certifications — these are third-party verified. No public status page found. No G2/Trustpilot reviews surfaced for the agent platform. Y Combinator-backed.

P3-B Scoped Permissions — 2/5
Pricing tiers limit agent count and message volume. Public/private agent status available. But no agent-scoped permission model with time/amount/action boundaries documented.

P3-C Audit Trail — 0/5
No machine-accessible audit log or transaction record for agent interactions found in documentation or any scraped page.

P3-D Behavioral Consistency — 1/5
SOC 2 certification implies some governance. No evidence of versioned terms, change notices, or explicit stability commitment found.

P4-A Friction-Free Activation — 4/5
Free tier available with up to 4 agents and unlimited conversations, no credit card required. Self-serve API key access documented. No human gate for entry-level access. Enterprise requires sales.

P4-B Agent Decision Signals — 2/5
Free tier provides a trial signal, and message/agent limits create a clear upgrade trigger. But no programmatic, agent-legible signals (webhooks on limit approach, status flags, decision metadata) documented.

P5-A Integration Depth — 3/5
1,500+ tool integrations and the ability to upload proprietary knowledge (PDF, DOCX, PPTX, EML, Notion, Drive, websites) creates meaningful data lock-in. Switching requires re-uploading knowledge and rebuilding tool configurations. Mild-to-moderate switching cost.

P5-B Agent Memory Layer — 2/5
Knowledge Context and Knowledge Tools enable agents to maintain embedded context with configurable retrieval thresholds. Account-level knowledge persists. But no dedicated agent memory API accessible programmatically — memory lives inside agent configuration, not a queryable layer.

P5-C Programmatic Renewal — 1/5
Standard SaaS subscription billing. No programmatic renewal API or agent-accessible subscription management endpoints found.

P5-D Compounding Value Signal — 1/5
Marketplace grows over time with more agents and tools. Fine-tuned agents accumulate usage history. But no agent-readable signals of compounding value (usage trends, improvement metrics, network effects) exposed programmatically.

Rubric v1 (April 2026). Scores reflect the company's state on the audit date and may have improved since.