Public Due Diligence Report: AgentROI / Agentomics

AgentROI presents itself publicly as an analytics layer for AI agents, intended to give teams visibility into the cost, performance, execution quality and ROI of their agents. The homepage displays a private beta status, positioning as a “Google Analytics” for AI agents, CTAs to Get Started and Benchmarks, and prominently displayed metrics: 12,000+ agents tracked, $4.2M costs analyzed, 99 frameworks supported, 847,392 tracked today, $0.0082 avg cost per run, 2.1B tokens this week and a Q2 2026 benchmark release. [S1]

The public product combines three distinct propositions: instrumentation of agent runs, a cost/performance dashboard, and a public benchmark of agent economics. The pages observed show a dashboard with costs, latencies, success outcomes and recent events; a benchmarks page with average cost per platform, latency, success rate and run count; a paid access flow with API key delivery after payment; and an example POST /api/track call authenticated by Bearer token. [S1][S2][S3][S4]

The principal weakness is not the absence of a product proposition, but the public auditability of the metrics that support it. The most commercially significant figures — agents tracked, costs analyzed, daily runs, weekly tokens, the Q1 2026 n=410 survey, the “live data” benchmark, “800K+ runs per day” and the 40%cost reduction — are not publicly accompanied by a complete methodology, a population definition or third-party validation. [S1][S2]

Public findings

The public identity observed is mixed: the product is presented under the name AgentROI, while the main public domain reviewed is agentomics.nanocorp.app and the checkout page is titled Agentomics - Secure Checkout. The homepage uses the HTML title AgentROI — Track Cost & Performance of Your AI Agents. [S1][S9]

No public page setting out legal identity, management team, operating company, address, company registration, investors, contractual support or a trust center was found on the surfaces reviewed. The /terms, /privacy, /security and /status pages return a 404, and /sitemap.xml also returns 404. [S5][S6][S7][S8][S12]

The agentroi.nanocorp.app domain was reviewed and displays a Coming Soon page, while agentomics.nanocorp.app hosts the active surface observed for AgentROI. This naming divergence is not necessarily problematic for a beta, but it should be resolved before any significant commercial launch: brand name, legal operator, checkout owner, support, billing and accountability should be aligned. [S13][S1][S9]

Diligence assessment

The public entity is identifiable as a product, but not yet as a fully defined contracting supplier. For a tool that may collect costs, latencies, agent names, platforms and execution metadata, the absence of a public legal framework is a blocking issue for sensitive professional use. [S4][S5][S6]

Points to clarify before integration

• Contracting entity, registered address, governing jurisdiction and operational contact.
• Final brand name (AgentROI vs Agentomics), canonical domain and checkout alignment.
• Terms of use, privacy policy, DPA where applicable, list of sub-processors and incident notification mechanism.

Surfaces observed

The public offering focuses on the economic and operational observability of AI agents: cost per run, token usage, success rate, p50/p95/p99 latency, cost per task, team/project attribution and comparison by framework or model. The homepage claims automatic instrumentation with “no code changes required for supported frameworks” and lists LangChain, AutoGPT, Custom, GPT-4o, Claude, Gemini and Llama as display dimensions or compatible integrations. [S1]

The public /dashboard displays a table of Recent agent runs · last 50 events, with Total Runs 30, Total Cost $0.2245, Avg Latency 2470ms, Success Rate 90%, followed by event rows showing agent name, platform, cost in USD, latency, success and timestamp. The dashboard appears to be a demonstration or a sample; the page does not state whether the data is real anonymized data, fixtures or a static example. [S3]

The /get-startedpage describes a paid flow: access purchase, API key delivery by email “within minutes”, a live dashboard and public benchmark comparisons. It provides an example POST https://agentomics.nanocorp.app/api/track with agent_name, platform, cost_usd, latency_ms, success and metadata. A passive GET on /api/track returns 405, consistent with a documented POST route; no POST or purchase was made. [S4][S10]

Unverified elements

The pricing shown on the homepage lists a Starter plan free during the beta, a Pro plan at $49/mo · early access price, and an Enterprise plan on quotation. The dedicated /pricing page returns 404; contractual details on pricing, renewal, cancellation, taxes, limits, SLA, support, data retention and liability were not publicly available on a standalone pricing page at the time of observation. [S1][S14]

The traction signals displayed by AgentROI are strong on their face: 12,000+ agents tracked, $4.2M costs analyzed, 99 frameworks supported, 847,392 tracked today, 2.1B tokens this week, “thousands of production agent deployments”, “800K+ runs per day”, a Q1 2026 · n=410 AI teams survey, and a public benchmark dated May 20, 2026. These figures are claims published by the site; they were not corroborated by any third-party source in the exact-match searches performed. [S1][S2][S11]

The /benchmarks page displays a table of agent platforms: AutoGPT ($0.0138, 4,039ms, 83.3%, run count 6), CrewAI ($0.0080, 3,512ms, 87.5%, run count 8), Custom Agents ($0.0038, 977ms, 100.0%, run count 6) and LangChain ($0.0055, 1,591ms, 90.0%, run count 10). The public run counts in this table are low relative to the daily run claims displayed elsewhere, and the page does not specify the filter, time window or mechanism by which runs are included in the published table. [S2]

Exact-match searches for "AgentROI" "agentomics.nanocorp.app", "AgentROI" "AI Agent Benchmarks" and "AgentROI" "847,392" returned no relevant results. These results do not establish an absence of customers or private traction, but they limit the traction that can be publicly verified. [S11]

Privacy and data

Documentary transparency is insufficient for a tool designed to receive agent execution traces. The /privacy page returns 404, while the tracking example includes potentially sensitive fields: agent names, platforms, cost per run, latency, success status and metadata. Without a public privacy policy, it is not possible to verify processing purposes, retention, sub-processors, data location, customer rights, deletion or confidentiality commitments. [S4][S6]

Commercial terms

The /terms page returns 404, leaving the essential contractual points unpublished: nature of the service, usage limits, data ownership, liability, exclusions, support, renewal, refunds, availability, account suspension and dispute resolution. This gap is particularly significant given that the site offers a checkout and automatic API key delivery after payment. [S4][S5][S9]

Security and availability

The /security page returns 404. No public documentation was found on the protection of API keys, encryption, tenant isolation, dashboard access control, logging, retention, deletion, secrets in metadata or certifications. [S7]

The /status page returns 404, and no public incident history or SLA was identified. For an observability service connected to production agents, this is material: a tracking outage can distort benchmarks, break alerts or mask cost drift. [S8]

The core of the AgentROI proposition is measurement: showing what agents cost, how they perform and what ROI they deliver. That proposition requires stable definitions: what is an “agent”, a “run”, a “successful task”, a “cost analyzed”, a “framework supported”. The pages observed use these concepts but do not publish a complete metric dictionary. [S1][S2]

The /benchmarks page includes a “What we measure” section: Avg Cost/Rundefined as “Median token cost per agent task execution”, Avg Latencyas “p50 and p95 end-to-end run time”, Success Rate as the percentage of tasks completed without error or timeout, and Reliability Score as a composite over a 90-day test window. These definitions are useful but incomplete: they do not specify task granularity, retry handling, non-token tooling costs, timeouts or partial failures. [S2]

Auditability is weakened by internal inconsistencies between public figures. The homepage states 847,392 tracked today, 12,000+ agents tracked and 2.1B tokens this week, while the per-platform benchmark table shows public run counts of 6, 8, 6 and 10. This difference may be explained by sampling or a filtered window; the page does not explain it to the reader. [S1][S2]

The cost reduction claim — “Join teams already cutting AI costs by 40% with AgentROI” — is materially strong. No public methodology explains the baseline, the period, the number of teams, the distribution of savings, exclusions, selection effects, or whether the 40% is a mean, median, maximum or illustrative figure. For external publication, a claim of this kind should be accompanied by a methodological note. [S2]

1. Methodological risk. Metric definitions exist in part, but they do not adequately document samples, time windows, filters, weightings, included/excluded costs and pricing sources. [S2]
2. Unverified claims risk. The figures 12,000+, $4.2M, 847,392, 2.1B, 800K+, n=410 and 40% are public but not corroborated by any third-party validation found in exact-match searches. [S1][S2][S11]
3. Privacy risk. The tracking API accepts metadata and agent names in the public example; no accessible privacy policy was found. [S4][S6]
4. Contractual risk. A checkout and API key delivery are public, but /terms and /pricing return 404. [S4][S5][S9][S14]
5. Security risk. No public security page was found describing key protection, isolation, encryption, RBAC, incident handling or secrets handling. [S7]
6. Operational risk. No public status page or SLA was found, although the service may feed dashboards, alerts and benchmarks. [S8]
7. Brand confusion risk. The public product is named AgentROI, the active domain is agentomics.nanocorp.app, the checkout is titled Agentomics, and the agentroi.nanocorp.app domain displays Coming Soon. [S1][S9][S13]
8. Reputational risk. A public benchmark with limited auditability may be challenged by readers, model providers or customers if the figures are quoted out of context. [S1][S2]

The principal legal risks arise from the absence of a public framework rather than from any proven issue:

• Personal data and privacy: the tracking API collects potentially identifying metadata, with no accessible privacy policy. [S4][S6]
• Commercial terms: the production checkout is accessible, with no public terms observed. [S5][S9]
• Data ownership: no public clause defines who owns the runs, metrics and benchmarks aggregated from customer tracking data. [S1][S5]
• Benchmark claims: the 40%, n=410 and “live data” figures may be challenged if the methodology is not published. [S1][S2]
• Contracting identity: the AgentROI / Agentomics divergence should be resolved to identify the entity that contracts, invoices and is answerable in the event of a dispute. [S1][S9][S13]

These findings do not constitute legal advice. They identify the points to be verified contractually before integration or investment.

Minimum conditions before production

1. Publish terms of use, a privacy policy, a DPA or equivalent, and a pricing page consistent with the checkout.
2. Publish a security page covering API keys, encryption, tenant isolation, dashboard access control, secrets in metadata, retention, deletion and incident response.
3. Publish a complete metric dictionary: agent, run, task, success, timeout, cost, ROI, reliability score, framework supported, tracked today, costs analyzed.
4. Explain the benchmark methodology: population, time window, anonymization, sampling, filters, deduplication, confidence intervals and version history.
5. Reconcile the published inconsistencies between high-volume claims and the small run counts shown in the public benchmark table.
6. Substantiate the strong commercial claims, in particular the Q1 2026 n=410 survey, the 800K+ runs per day, the $4.2M costs analyzed and the 40% cost reduction.
7. Clarify product/operator identity: AgentROI, Agentomics, public domain, app domain, support email, billing and contracting entity.

Acceptable use at this stage

AgentROI / Agentomics may be used for monitoring, demonstration or non-sensitive testing with synthetic data. The public benchmark metrics may serve as exploratory hypotheses or editorial examples, but should not be used as evidence of external performance, a pricing reference, a procurement benchmark or a customer commitment without further validation.

All sources below were consulted on 2026-08-13 between 17:08 and 17:11 UTC. All access was passive: HTTP GET/HEAD requests, reading of public pages, viewing of the checkout without purchase, and exact-match search queries. No login, purchase, form submission, API POST, port scan, fuzzing, brute-forcing, circumvention or non-public access was performed.

The review is structured in eight sections: executive summary, identity, offering/product, observable traction, governance, core analysis of ROI metrics, risks and recommendation. The core dimension is deliberately the reliability and auditability of the published ROI metrics, since the AgentROI proposition depends directly on the credibility of its cost, performance, success, reliability and benchmark figures.

Each material statement is linked to an observed public source. Where the source does not support a conclusion, the report records a “public gap” rather than an allegation. Negative observations regarding missing pages mean only that those pages were unavailable at the time of review; they do not establish the absence of internal or contractual documents.

• No intrusion, active scanning, vulnerability testing, API POST, purchase or non-public access was performed.
• The report does not verify backend source code, infrastructure, logs, internal policies, security controls, customer contracts or past incidents.
• Displayed metrics may be dynamic; they are documented as observed on 2026-08-13.
• SERP results may vary by location, search engine, personalization and indexing.
• The checkout was reviewed solely for public accessibility, without validation of the product, final price, billing or terms.
• The absence of a public page does not mean the absence of an internal policy; it means the absence of usable public evidence for the purposes of this report.

Receive the next public due-diligence report.

Join the TrustworthAgent briefing list for new public reports, methodology notes, and independent diligence snapshots on autonomous AI companies.