Frequently Asked Questions
Common questions about using AWI, integrating with your agents, and understanding source assessments.
General
AWI (Agentic Web Index) is measurement infrastructure for AI agents. It measures the delta between how web systems present themselves and how they actually behave — so agents can make informed access decisions before consuming content.
Sign up for a free account to get an API key. You can query the API directly, connect via MCP, or install the Claude Code plugin. Free accounts get 25 lookups and 5 assessments per day.
Any AI agent or application that accesses web content can use AWI. Common use cases include pre-access trust checks in autonomous agents, source routing for research workflows, content assessment, and building trust-aware agent pipelines.
API & Integration
AWI offers a REST API, a Streamable HTTP MCP endpoint (compatible with any MCP client), a Python SDK, and a Claude Code plugin. All integration code is MIT-licensed.
Yes. You can integrate AWI into commercial products, SaaS applications, and paid agent workflows. The API is a hosted service — use it however you need. The only restriction is building a competing hosted source intelligence service.
Free accounts are limited to 25 lookups and 5 assessments per day. Upgrade to Pro for unlimited lookups and assessments with a flat $5/month subscription.
AWI returns composite scores, concern level (none/low/moderate/high/critical), access status, and five dimensional assessments covering agent-control risk, content integrity, operational accessibility, provenance, and observed outcomes for each source. Pro accounts get the full dimensional breakdown.
Scoring & Methodology
AWI uses a 4-layer analysis architecture: content acquisition, pattern matching, structural analysis, and behavioral testing — combined with longitudinal analyzers. Scores are pre-computed and updated based on demand-driven re-evaluation.
Source owners can improve scores by publishing .well-known/awi.json manifests, maintaining clean content, and supporting agent-friendly access patterns. Agents can report outcomes that feed into the OBS (Observed Interaction Success) dimension.
Sources with high interaction frequency are evaluated near-real-time, moderate frequency daily, and low frequency periodically. This demand-based approach ensures active sources maintain current scores without unnecessary crawling.
Open Source & Licensing
The Python SDK, Claude Code plugin, and open specifications are MIT-licensed and available at github.com/agenticwebindex/awi. The API and MCP endpoint are provided as hosted services.
The API and scoring engine are provided as a hosted service. If you have specific self-hosting requirements, contact us to discuss options.
File issues or contribute to the SDK and plugin at github.com/agenticwebindex/awi. For API issues or feature requests, contact [email protected].
Still have questions?
We're happy to help with integration questions or custom requirements.