AI agents are increasingly the first step in software evaluation — researching vendors, comparing plans, and shortlisting tools before a human gets involved. But most SaaS companies weren't built with agents in mind. They lack machine-readable APIs, structured pricing pages, and the protocol support agents need to act autonomously.
This index scores 510+ B2B SaaS vendors on how well they support AI agents today. Each score reflects 12+ automated probes across five dimensions: whether agents can crawl the site, call the API without human help, understand the pricing, parse what the company does, and plug into agent tooling like MCP servers and SDKs.
Higher scores mean an agent is more likely to discover, understand, and successfully use that vendor — which matters as AI-driven procurement becomes standard. How scoring works →
Each vendor is scored using automated probes that run against their public web presence — no login required, no vendor input needed. We check the same things an AI agent would check when it encounters a vendor for the first time: can it read the site, find the API, understand the pricing, and plug into agent tooling?
Raw probe results are converted to points across five dimensions, then normalized to a 0–100 score. A score of 70+ means the vendor is well-positioned for agent-driven workflows. Below 40 means significant gaps that would prevent most agents from using the product autonomously.
Scores are computed independently by Proven. Vendors cannot pay to improve their ranking — only fixing the underlying technical gaps moves the score. Full methodology →
Scores are refreshed periodically as vendors make changes to their products. The last-scored date is visible on each vendor's detail page. Vendors who have made improvements can request a rescore using the button above — rescores typically run within a few business days.
All probes run against publicly accessible URLs — no login, no vendor cooperation required. This means private APIs, gated documentation, and features behind authentication are not scored. A vendor with excellent private-API tooling but no public-facing signals will score lower than their actual agent-readiness warrants.
MCP and agent ecosystem probes rely on community registries (modelcontextprotocol/servers, punkpeye/awesome-mcp-servers, Context7). Listings that exist but haven't been submitted to these registries won't be detected. The scoring reflects what an agent would find, not what exists internally.
Each vendor detail page includes a personalized improvement plan — unlock it by entering your email. The plan ranks fixes by points gain and effort, so you can prioritize the highest-leverage changes. Common quick wins include publishing an llms.txt file, allowing AI bots in robots.txt, adding a schema.org JSON-LD block to the homepage, and publishing an OpenAPI spec.
Proven also works directly with VC-backed companies to accelerate agent readiness improvements as part of its vendor marketplace. Learn more about Proven →
Proven is a marketplace used by VC firms and their portfolio companies to access deals, evaluate tools, and manage vendor relationships. As AI agents take on more of the procurement research process, the vendors that agents can actually read, compare, and use will have a structural advantage in being discovered and shortlisted.
The Agent Report Card is Proven's public contribution to measuring that readiness — open data for builders, investors, and procurement teams who want to understand which parts of the SaaS landscape are genuinely ready for the agentic era.
An AI agent readiness score measures how well a software vendor's product can be discovered, read, and used by AI agents operating autonomously — without human intervention at each step. It covers whether the vendor has machine-readable APIs, structured documentation, transparent pricing, and support for agent protocols like MCP.
Each vendor is tested across five dimensions using 12+ automated probes: Crawlability (can agents reach the site?), API Readiness (is there a public OpenAPI spec and docs site?), Pricing Clarity (can agents read pricing without guessing?), Structured Data (schema.org markup for agent comprehension), and Agent Ecosystem (MCP servers, SDKs, LangChain integrations, auth friendliness, webhooks). Raw probe points are normalized to a 0–100 scale.
The rankings are useful for: (1) AI builders evaluating which vendors to integrate with, (2) procurement teams that want to know which tools support autonomous agent workflows, and (3) software vendors who want to understand and improve their own agent-readiness posture.
Scores are refreshed periodically as vendors ship changes. Vendors can also request a rescore using the 'Request a ranking' button if they've made improvements since their last score.
Most B2B SaaS companies were built before AI agents became a meaningful distribution channel. The capabilities that matter to agents — public OpenAPI specs, machine-readable pricing, webhook docs, MCP servers — are still rare. Only about 1% of vendors in our index score 70 or above.
Yes. Use the 'Request a ranking' button on the marketplace to submit your domain. If your company is already listed and you believe a score is inaccurate, email us at hello@proven.email with the specific probe you think is wrong and the evidence.
Proven gives venture-backed startups exclusive access to deals from 900+ vendors.
Drop your company's domain and your work email. We'll score how well AI agents can find and parse your site, then email you the report. Takes ~30 seconds.
Want the full breakdown of all 7 checks, plus a personalized improvement plan? Drop your email below.
See full report → drop your email