# Research & Analysis — The Agents Index

> Source: https://theagentsindex.com/blog (curated, sourced)

Data-driven research and analysis on AI agents, frameworks and agentic tools, sourced reports built to be cited.

60 reports. Each links to its answer-first markdown twin.

- [Data residency: a census of 145 listings](https://theagentsindex.com/blog/data-residency-census.md) (2026-10-05) — 34 of 145 listings answer yes for “Data residency”. Counted from our own datasheet, with the query that reproduces every figure.
- [Runs unattended: a census of 145 listings](https://theagentsindex.com/blog/autonomous-census.md) (2026-10-04) — 99 of 145 listings answer yes for “Runs unattended”. Counted from our own datasheet, with the query that reproduces every figure.
- [Claude models: a census of 145 listings](https://theagentsindex.com/blog/claude-census.md) (2026-10-03) — 87 of 145 listings answer yes for “Claude models”. Counted from our own datasheet, with the query that reproduces every figure.
- [Ships it: a census of 145 listings](https://theagentsindex.com/blog/ships-code-census.md) (2026-10-01) — 18 of 145 listings answer yes for “Ships it”. Counted from our own datasheet, with the query that reproduces every figure.
- [Multi-agent: a census of 145 listings](https://theagentsindex.com/blog/multi-agent-census.md) (2026-09-30) — 81 of 145 listings answer yes for “Multi-agent”. Counted from our own datasheet, with the query that reproduces every figure.
- [API access: a census of 145 listings](https://theagentsindex.com/blog/api-access-census.md) (2026-09-29) — 83 of 145 listings answer yes for “API access”. Counted from our own datasheet, with the query that reproduces every figure.
- [Free to use: a census of 145 listings](https://theagentsindex.com/blog/free-to-use-census.md) (2026-09-28) — 84 of 145 listings answer yes for “Free to use”. Counted from our own datasheet, with the query that reproduces every figure.
- [Pooled budget: a census of 145 listings](https://theagentsindex.com/blog/pooled-budget-census.md) (2026-09-27) — 67 of 145 listings answer yes for “Pooled budget”. Counted from our own datasheet, with the query that reproduces every figure.
- [Done-for-you service: a census of 145 listings](https://theagentsindex.com/blog/agency-service-census.md) (2026-09-26) — 12 of 145 listings answer yes for “Done-for-you service”. Counted from our own datasheet, with the query that reproduces every figure.
- [Writes code: a census of 145 listings](https://theagentsindex.com/blog/writes-code-census.md) (2026-09-25) — 51 of 145 listings answer yes for “Writes code”. Counted from our own datasheet, with the query that reproduces every figure.
- [On the command line: a census of 145 listings](https://theagentsindex.com/blog/surface-cli-census.md) (2026-09-24) — 67 of 145 listings answer yes for “On the command line”. Counted from our own datasheet, with the query that reproduces every figure.
- [Certifications: a census of 145 listings](https://theagentsindex.com/blog/compliance-census.md) (2026-09-23) — 63 of 145 listings answer yes for “Certifications”. Counted from our own datasheet, with the query that reproduces every figure.
- [Model choice: a census of 145 listings](https://theagentsindex.com/blog/model-agnostic-census.md) (2026-09-22) — 75 of 145 listings answer yes for “Model choice”. Counted from our own datasheet, with the query that reproduces every figure.
- [MCP server: a census of 145 listings](https://theagentsindex.com/blog/mcp-access-census.md) (2026-09-21) — 71 of 145 listings answer yes for “MCP server”. Counted from our own datasheet, with the query that reproduces every figure.
- [Seat model: a census of 144 listings](https://theagentsindex.com/blog/seat-model-census.md) (2026-09-20) — 48 of 144 listings answer yes for “Seat model”. Counted from our own datasheet, with the query that reproduces every figure.
- [Whole-codebase work: a census of 145 listings](https://theagentsindex.com/blog/codebase-scale-census.md) (2026-09-19) — 14 of 145 listings answer yes for “Whole-codebase work”. Counted from our own datasheet, with the query that reproduces every figure.
- [Opt out of training: a census of 145 listings](https://theagentsindex.com/blog/training-optout-census.md) (2026-09-18) — 69 of 145 listings answer yes for “Opt out of training”. Counted from our own datasheet, with the query that reproduces every figure.
- [Admin controls: a census of 145 listings](https://theagentsindex.com/blog/admin-policy-census.md) (2026-09-16) — 61 of 145 listings answer yes for “Admin controls”. Counted from our own datasheet, with the query that reproduces every figure.
- [Open models: a census of 145 listings](https://theagentsindex.com/blog/open-models-census.md) (2026-09-15) — 54 of 145 listings answer yes for “Open models”. Counted from our own datasheet, with the query that reproduces every figure.
- [Agent permissions: a census of 145 listings](https://theagentsindex.com/blog/agent-permissions-census.md) (2026-09-14) — 65 of 145 listings answer yes for “Agent permissions”. Counted from our own datasheet, with the query that reproduces every figure.
- [Debugs and fixes: a census of 145 listings](https://theagentsindex.com/blog/debugs-code-census.md) (2026-09-13) — 34 of 145 listings answer yes for “Debugs and fixes”. Counted from our own datasheet, with the query that reproduces every figure.
- [Writes tests: a census of 145 listings](https://theagentsindex.com/blog/writes-tests-census.md) (2026-09-12) — 20 of 145 listings answer yes for “Writes tests”. Counted from our own datasheet, with the query that reproduces every figure.
- [Bring your own key: a census of 145 listings](https://theagentsindex.com/blog/byok-census.md) (2026-09-10) — 55 of 145 listings answer yes for “Bring your own key”. Counted from our own datasheet, with the query that reproduces every figure.
- [Reviews code: a census of 145 listings](https://theagentsindex.com/blog/reviews-code-census.md) (2026-09-09) — 26 of 145 listings answer yes for “Reviews code”. Counted from our own datasheet, with the query that reproduces every figure.
- [In your pipeline: a census of 145 listings](https://theagentsindex.com/blog/surface-ci-census.md) (2026-09-08) — 28 of 145 listings answer yes for “In your pipeline”. Counted from our own datasheet, with the query that reproduces every figure.
- [Open source: a census of 145 listings](https://theagentsindex.com/blog/open-source-census.md) (2026-09-07) — 30 of 145 listings answer yes for “Open source”. Counted from our own datasheet, with the query that reproduces every figure.
- [Self-hosted: a census of 145 listings](https://theagentsindex.com/blog/self-hosted-census.md) (2026-09-06) — 33 of 145 listings answer yes for “Self-hosted”. Counted from our own datasheet, with the query that reproduces every figure.
- [In your editor: a census of 145 listings](https://theagentsindex.com/blog/surface-ide-census.md) (2026-09-05) — 33 of 145 listings answer yes for “In your editor”. Counted from our own datasheet, with the query that reproduces every figure.
- [Which AI Support Agents Skip the Sales Call? 4 of 9](https://theagentsindex.com/blog/ai-support-agent-self-serve-2026.md) (2026-08-25) — We checked all 9 support agents in our index for self-serve access. Four publish a price and a signup page; the other five start with a sales call.
- [25 AI Coding Agents Compared: Pricing and API Access](https://theagentsindex.com/blog/coding-agents-compared-2026.md) (2026-08-25) — A field-by-field comparison of all 25 coding agents in our directory, split by pricing model and whether each one exposes a public API.
- [Which of 14 AI Agent Frameworks Locks You to One Model?](https://theagentsindex.com/blog/agent-framework-model-lockin-2026.md) (2026-08-23) — We checked model-provider support across our 14 indexed agent frameworks. Only the Claude Agent SDK locks you to a single model vendor by design.
- [Which AI Agents Are Free to Start? 60 of 94](https://theagentsindex.com/blog/free-tier-across-ai-agents-2026.md) (2026-08-19) — We read the published pricing tiers of the 94 classified AI agents in our index on 27 August 2026. 60 offer an ongoing free tier and 34 do not.
- [Research or Data Agent? 6 of 11 Return Structured Data](https://theagentsindex.com/blog/research-agent-or-data-agent-2026.md) (2026-08-14) — Our Research & data agents category holds 11 tools. Re-verified 26 August 2026: 6 return structured, machine-readable output as a named feature; 5 return prose.
- [Which AI Agents Have a Public API? 72 of the 81 We Could Confirm](https://theagentsindex.com/blog/ai-agent-api-census-2026.md) (2026-08-13) — We checked all 94 AI agents in our index for a public API on 22 August 2026. 72 of the 81 we could determine offer one, and 34 of the 35 you can self-host.
- [AI Agent Entry Prices: 40 of 92 Publish One, and the Median Is $20](https://theagentsindex.com/blog/ai-agent-entry-price-census-2026.md) (2026-08-13) — We read the pricing table of all 92 tools in our index on 18 August 2026. 40 publish an exact flat monthly price in dollars, and the median is $20 a month.
- [CLI, Web App or Library? An Interface Census of 93 AI Agents](https://theagentsindex.com/blog/ai-agent-interface-census-2026.md) (2026-08-13) — We recorded the primary interface of 93 of the 94 AI agents in our index on 20 August 2026. All 14 library-first tools self-host; 4 of 37 web apps do.
- [AI Agent Pricing Models: 21 of 93 Tools Bill by Usage](https://theagentsindex.com/blog/ai-agent-pricing-model-census-2026.md) (2026-08-13) — We classified the pricing meter on 93 of the 94 AI agents published in our index on 26 August 2026. 21 bill by usage, 15 of them voice or support agents.
- [How Many AI Agent Tools Publish a Full Price List? 30 of 92](https://theagentsindex.com/blog/ai-agent-pricing-transparency-census-2026.md) (2026-08-13) — We read all the pricing tiers recorded for the 93 tools in our index on 25 August 2026. 30 publish a figure for every tier, and 8 publish no price anywhere.
- [AI Agents for Sales & Marketing: What the 7 in Our Index Actually Do](https://theagentsindex.com/blog/ai-sales-marketing-agents-2026.md) (2026-08-13) — The 7 sales and marketing agents we index split three ways: outbound, inbound, and the account-data layer underneath. Only 2 let you start without a sales call.
- [GitHub Copilot's Coding Agent: Three Surfaces, Three Defaults](https://theagentsindex.com/blog/github-copilot-coding-agent-2026.md) (2026-08-13) — Copilot's coding agent is the cloud one: you assign it an issue, it opens a pull request. It is not agent mode in VS Code, and it does not ask before it acts.
- [Which Coding Agents Run Commands Without Asking? 21 Tools, Checked](https://theagentsindex.com/blog/coding-agent-autonomy-defaults-2026.md) (2026-08-11) — Cursor auto-applies edits but asks before commands. Codex CLI and OpenCode run both by default. We checked all 21 coding agents in our index against their own docs.
- [Which AI Agent Frameworks Support A2A? 12, Verified](https://theagentsindex.com/blog/a2a-protocol-support-frameworks-2026.md) (2026-08-09) — A2A lets agents on different frameworks talk to each other. We checked all 12 frameworks in our index: 4 ship it natively, 3 as an add-on, 4 don't support it yet.
- [Does Your Coding Agent Train on Your Code? 21 Tools, Checked](https://theagentsindex.com/blog/coding-agent-data-training-defaults-2026.md) (2026-08-06) — SOC 2 confirms an audit was passed, not whether your code trains a model. We checked the default training policy, tier by tier, for 21 coding agents in our index.
- [Why AI Support-Agent 'Resolution Rate' Claims Aren't Comparable](https://theagentsindex.com/blog/support-agent-resolution-rate-claims-2026.md) (2026-08-05) — Seven AI support agents in our index report success on different metrics: resolution, containment, deflection, accuracy. Here's what each one measures, and hides.
- [What 'Multi-Agent' Actually Means: Six Shapes, 36 Tools](https://theagentsindex.com/blog/what-multi-agent-actually-means-2026.md) (2026-08-05) — 36 of the 95 listings in The Agents Index carried our multi-agent tag on 28 August 2026. We sorted the sourced facts into six different shapes.
- [Terminal-Bench 2.1: How Our Indexed Coding Agents Actually Score](https://theagentsindex.com/blog/terminal-bench-coding-agent-scores-2026.md) (2026-08-02) — Stanford/Laude Institute's independent Terminal-Bench results, matched against the coding agents in our index: accuracy, model pairing and cost, not vendor claims.
- [GitHub Star Velocity: 30 Open-Source AI Agents Measured 7 Days Apart](https://theagentsindex.com/blog/github-star-velocity-open-source-ai-agents-2026.md) (2026-08-01) — Stars for 32 trackable open-source AI agents in our index, counted 2026-08-18 and 2026-08-25. OpenAI Codex CLI gained 11,317 stars in 7 days.
- [Which AI Agents Can You Self-Host? A Census of 94 Tools](https://theagentsindex.com/blog/ai-agent-deployment-census-2026.md) (2026-07-28) — Of the 94 AI agents in our index, 42 can run on infrastructure you control and 52 are cloud-only. Support and sales agents offer zero self-hosting.
- [Are AI Coding Agents & Frameworks Actually Open Source?](https://theagentsindex.com/blog/open-source-license-census-ai-agents-2026.md) (2026-07-24) — We read the license behind every AI agent in our index. All 94 product listings now have a determined answer, and 33 are open source in some form.
- [Which AI Agents Won't Tell You What LLM Powers Them?](https://theagentsindex.com/blog/model-transparency-census-2026.md) (2026-07-23) — We checked all 93 AI agents in our index for the LLM underneath. Of the 92 we could classify, 23 name no model at all, and it tracks how the tool is sold.
- [How Much Do AI Agents Actually Cost? A Pricing FAQ](https://theagentsindex.com/blog/ai-agent-pricing-faq-2026.md) (2026-07-22) — Intercom Fin, n8n, Clay, Lindy and LangChain each price agents differently: per resolution, per credit, per seat. What each one actually charges.
- [What Is MCP, and Which AI Agents Mention It?](https://theagentsindex.com/blog/mcp-support-across-ai-agents-2026.md) (2026-07-22) — We searched each listing's own attributes, key features and integrations for the text "MCP" on 27 August 2026: 70 of 94 mention it, 24 do not.
- [9 AI agent tools that changed how they charge in 2026](https://theagentsindex.com/blog/agent-tool-pricing-shifts-2026.md) (2026-07-21) — Nine 2026 pricing-change case studies, with verified old and new terms and what they mean for buyers of “free,” “bundled,” or “unlimited” plans.
- [Who owns the AI agent you're building on? A 2025–2026 field guide](https://theagentsindex.com/blog/ai-agent-acquisition-wave-2025-2026.md) (2026-07-21) — Six tools in our index changed hands or raised historic rounds in nine months. What happened, what it already broke, and how to judge the risk before you adopt.
- [AI Agent Statistics 2026: Adoption, Production, and Trust](https://theagentsindex.com/blog/ai-agent-statistics-2026.md) (2026-07-21) — Sourced 2026 statistics on AI agent adoption, production and developer trust, from Gartner, McKinsey, LangChain and Stack Overflow, plus our own 94-tool index.
- [AI voice agent funding vs. real growth: the 2026 gap](https://theagentsindex.com/blog/ai-voice-agent-funding-vs-growth-2026.md) (2026-07-21) — Vapi, Retell AI, Bland AI, ElevenLabs and Cartesia raised very different amounts. The one that raised the least reports the fastest, most profitable growth.
- [Using LangChain tools inside CrewAI agents: a working guide](https://theagentsindex.com/blog/langchain-tools-inside-crewai-integration-guide.md) (2026-07-21) — CrewAI and LangChain are usually pitched as rivals. Their own docs disagree: the verified, working pattern for wrapping a LangChain tool for a CrewAI agent.
- [The State of AI Agents 2026: A Census of 93 Researched Tools](https://theagentsindex.com/blog/state-of-ai-agents-2026.md) (2026-07-20) — License, API, deployment and price for all 93 AI agents in our index, measured 26 August 2026. 33 are open source, and every self-host-only tool is one of them.
- [Terminal AI coding agents: open, portable, or locked in?](https://theagentsindex.com/blog/terminal-ai-coding-agents-open-portable-locked.md) (2026-07-16) — We verified the license, model portability and pricing of six terminal AI coding agents: Claude Code, Codex CLI, Gemini CLI, Aider, OpenCode and Goose.
- [How The Agents Index researches AI agents](https://theagentsindex.com/blog/how-we-research-ai-agents.md) (2026-07-13) — Our method for deciding what an AI agent actually does, what it costs, and whether it ships: sourced, quality-gated, and re-checked on a schedule.