AI Directory

The default answer when someone needs a capable form and does not want a pricing conversation. A thousand free responses a month clears most small business volume outright, and $15 a month buys the next two thousand. Recommended on value and breadth, not on intelligence: every branch is one you drew, so it collects what you thought to ask and nothing more. Reach for something model-driven only when the follow-up question depends on the answer.

Best for

Our recommended pick for conversational intake. It is the most complete implementation of the idea that intake should end in a typed record rather than a transcript: you declare the fields and what a complete answer looks like, the agent asks in ordinary language, follows up where an answer is too vague to act on, accepts several answers in one sentence, and reads an uploaded licence or policy instead of making someone retype it. The record it returns is typed, validated and deduplicated, and it goes out through a signed webhook, Zapier, Make, n8n, a spreadsheet or an inbox. It runs as an embedded widget or a full conversation page, with its own theme, traits, knowledge and access rules per agent, and it covers the whole pipeline from first contact through qualification, quoting, onboarding, feedback and recurring check-ins. Over 250 Gnosaris are running for real businesses, and it is the only tool in this category Neomanex operates in production, which we declare. Right when answers arrive in fragments and someone is currently chasing the gaps by hand. Wrong when you need payments, signatures or generated documents, which is form builder territory.

Best for

The best workflow automation tool for teams that want full control, AI integration, and no per-workflow pricing.

Best for

The default choice for local LLM inference — and for good reason. 176K GitHub stars, 9M+ users, MIT license, one-command setup. Runs on any hardware from Raspberry Pi 5 to dual H100s, with a model catalog spanning 4,500+ open-weight options. The OpenAI-compatible API with streaming, tool calling, structured outputs, and embeddings means no vendor lock-in — swap from Ollama to OpenAI (or vice versa) by changing one URL. Ollama Cloud (Pro $20/mo, Max $100/mo) extends the same surface to managed inference. The recent $65M raise confirms sustained investment. For any developer who wants to run AI locally — whether for privacy, cost control, or offline use — Ollama is where you start.

Best for

OpenAI Codex is the most fully-realized autonomous coding agent in 2026 — hand it a ticket, walk away, and come back to a finished PR running in a cloud sandbox with tests passing. Unlike inline tools, Codex handles long-running parallel cloud tasks across multiple repos simultaneously. It trails Claude Code for interactive terminal work and Copilot for fast inline suggestions, so it's not a universal replacement. Codex is the right pick for developers who want AI that finishes whole tasks autonomously, not just suggestions while typing.

Best for

The best value in form building by a wide margin, and the right default when someone asks for "a form" without further qualification. Unlimited submissions on the free plan with logic, webhooks and integrations included is not a trial shape, it is the product. Pro at $24/mo buys branding removal, custom domains and drop-off analytics. It is still a form: fixed question order, hand-wired branching, and no ability to read a free-text answer and decide what to ask next. Recommended because within the thing it is, almost nothing is held back from you.

Best for

A promising open-source alternative to Zapier with strong developer experience, but the integration library and community are still maturing.

Best for

The best tool for browser-based automation and web scraping, but limited for backend or server-side workflows.

Best for

Bifrost is a free, Apache 2.0 AI gateway from Maxim that puts 20+ model providers behind one OpenAI-compatible API, with virtual-key budgets, semantic caching, fallbacks and an MCP gateway. Conditional on hardening: management authentication is off by default, and JFrog disclosed two unauthenticated code-execution flaws that rely on it, CVE-2026-90898 (CVSS 9.8, fixed in 2.1.0) and CVE-2026-86242 (CVSS 8.1, fixed in 2.0.0). Right for platform teams that enable auth and patch.

Best for

Claude Code is the most architecturally complete agentic coding tool we track: dynamic workflows, maker-checker sub-agents, cloud routines and a separate verifier model all ship in one engine across terminal, IDE, desktop and web. The condition is cost. Anthropic's own docs put the average at about $13 per developer per active day, or $150-250 per month, and agent teams use roughly 7x the tokens in plan mode. Right for teams that meter and cap spend.

Best for

ConvoForm is the only Apache 2.0 conversational form builder in this directory whose code you can run on your own infrastructure with Docker, and its hosted version is free with no card. Conditional because its stack names OpenAI for the AI layer and Clerk for sign-in, and the repository does not document where answers go when self-hosted, so verify the data path before relying on residency. It is a small project with about 100 stars and no support commitment we could find: budget engineering time, not a subscription.

Best for

DeepSeek's Search and Fetch API gives agents web search and page fetch for $0.75 per 1,000 successful requests each, with empty results and failures not billed. Conditional: the guide we read states no data-retention or privacy terms and no response-length limits, and calls through DeepSeek Harness, Claude Code or Codex stop being free at 00:00 Beijing time on November 1, 2026. It suits teams already on the DeepSeek API that want metered search without another vendor. We have not tested result quality.

Best for

Docker Sandboxes is a free way to let coding agents such as Claude Code, Codex or Gemini run with full sudo inside a microVM with its own kernel, Docker Engine, deny-by-default networking and proxy-injected credentials. Conditional on patching and mount discipline: releases from 0.28.0 until the 0.42.0 fix carry a critical macOS host-file escape, and the default mount is read-write, so the agent edits your working tree in place. Right for developers running agents unattended.

Best for

ElevenAgents is ElevenLabs' platform for voice, chat and email agents in 90+ languages, with a 15-minute free plan and $0.08 per additional call minute. Conditional because the LLM is billed separately on top at a model-dependent rate that the pricing page does not fold into the per-minute price, so the all-in per-minute cost is not stated there. It suits teams wanting telephony, channels and testing from one vendor. We have not tested it.

Best for

Worth a look specifically when volume is your problem: an uncapped free plan is rare, and a high-traffic form that would cost real money elsewhere costs nothing here. The condition is the five-form ceiling on free, which bites fast for anyone running campaigns rather than one form. Judge it on the form count, not the response count. And treat the conversational mode as a display style, not an intelligence: the questions are still fixed in the order you wrote them.

Best for

GitHub Copilot is the most widely deployed and IDE-flexible AI coding assistant in 2026 — the $10/month Pro tier delivers unlimited autocomplete and solid agent mode, exceptional value for augmentation-style workflows. But the June 2026 switch to credit-based billing transformed pricing for agent-heavy users: sessions can burn $30-40 in credits, with some users reporting up to 27x cost increases. Copilot remains the pragmatic default for developers who want AI in their existing editor without switching ecosystems.

Best for

GitLab Duo Agent Platform is the natural agent layer for teams already running GitLab: event-triggered flows and CI, security and review agents live where the code and pipelines are. The deciding caveat is cost and trust. Usage is metered in credits, the included allowance is promotional, and a critical CVSS 9.9 flaw hit self-hosted AI Gateways. Right for GitLab shops that patch fast; skip it if your code lives elsewhere.

Best for

The right answer when the form has to do more than collect answers: take a payment, capture a signature, run an approval chain, emit a PDF, or sit under a HIPAA requirement. Few competitors cover that span at all. The condition is that you are buying a large, busy platform and its complexity, and the free tier is genuinely small at 5 forms and 100 submissions a month with a 500 total submission ceiling. If you only need questions answered, lighter tools do it more cleanly for less.

Best for

LiteLLM is a strong free default for putting many model providers behind one OpenAI-format API on your own infrastructure, with virtual keys, budgets, load balancing and guardrails in the open-source plan. Conditional on security discipline: a PyPI supply-chain compromise, an MCP authentication bypass exploited in the wild, and on September 30, 2026 a critical advisory (CVSS 9.9) that lets an internal user forge proxy admin credentials. Right for platform teams that pin versions, patch within days and harden deployments.

Best for

A strong Zapier alternative with better pricing and more flexible workflow design, but fewer integrations and a steeper learning curve.

Best for

Meta Muse is the most consumer-ready personal AI agent yet: free to start, live on web, iOS, Android and WhatsApp, and willing to send email, book travel and buy things on your behalf. Conditional because every safety assurance behind those powers is Meta's own and unaudited, and we have not watched it execute a single purchase. Right for US consumers who will supervise an agent; wrong for anyone needing audited security over payments or health data.

Best for

Microsoft Execution Containers is a free, MIT-licensed containment layer: you declare in one JSON policy what an agent may touch, and a process, session, WSL or microVM backend enforces it on Windows 11, macOS and Linux. Conditional because the microVM backend is experimental, the session and WSL backends run only on Windows 11, and Claude Code support is only planned. The evidence we read is Microsoft's own announcement and repository, with no independent security assessment. It suits teams embedding agents in products.

Best for

Midjourney remains the strongest AI image generator for artistic quality and cinematic aesthetics — V8.2 delivers best-in-class photorealism and prompt adherence. However, it ships no C2PA content credentials or provenance watermark, putting it in active non-compliance with California SB 942 and the EU AI Act as of August 2, 2026, exposing commercial users to regulatory risk. For provenance-required workflows, use Adobe Firefly, DALL-E 3, or Google Imagen instead.

Best for

OpenShell is NVIDIA's free, Apache-2.0 runtime that sandboxes autonomous agents with kernel-level file, syscall and network policy, injects credentials only for approved endpoints, and checks policy changes before you approve them. Conditional because it is at 0.1.x, Windows runs through WSL 2 and is marked experimental, and anonymous telemetry is on by default. Right for teams that want to trial policy-enforced agent isolation on Linux or Apple Silicon; we have not tested it.

Best for

OpenAI Dots are ChatGPT agents that keep working toward your goals around the clock on their own cloud computer, with 4,000+ connectable apps and a computer you can open to inspect their work. The first dot is included in Pro and Business Premium. Conditional because this is a rolling launch limited to eligible markets, extra dots are only promised and unpriced, and extended limits cover just the first month. Right for Pro and Business Premium subscribers who want a persistent agent; we have not tested it.

Best for

OpenClaw is a credible free way to run a personal AI agent on your own hardware: MIT licensed, no paid tier, reachable from 29 chat channels, and able to browse, read and write files and run shell commands. Conditional because that reach is the risk: its docs disclaim it as a boundary between mutually adversarial users, and Unit 42 found malicious ClawHub skills, warning that installing a skill grants complete control over the agent's identity. Right for technical users who will harden and audit it; wrong for unmanaged shared deployments.

Best for

OpenCode is a free, MIT-licensed coding agent for terminal, desktop, IDE and web that works with 75+ model providers, plus optional paid Zen and Go model plans. The condition is trust: its permissions docs describe permission rules, not a sandbox, most permissions default to allow, and a High-severity advisory hit npm-, pnpm- and Bun-installed copies running `opencode serve`. Right for developers who patch fast and isolate it in a VM or container.

Best for

OpenRouter is the de-facto standard LLM gateway: one OpenAI-compatible API exposing 500+ models across 80+ providers, with unified billing, automatic failover, and no lock-in. It's an excellent default for model comparison and multi-provider routing. The conditions: a ~5.5% platform fee makes direct provider access cheaper at single-model scale, and Stripe agreed to acquire OpenRouter on August 19, 2026 (agreed, not closed, no price disclosed), leaving routing neutrality an on-record pledge rather than a structural guarantee. Best for developers and startups who value flexibility over the lowest per-token price.

Best for

The best pick when the words around the question matter as much as the question, which makes it strong for landing pages, bookings and paid order forms. The condition is the free tier: 30 submissions a month is a demo allowance, not an evaluation, so budget from the first real deployment. The annual submission accounting also rewards steady volume and punishes spiky months, so check your traffic shape against the tier before buying.

Best for

Perplexity Computer is a credible general-purpose agent for long, multi-step work: it turns a stated outcome into subtasks run by sub-agents in an isolated cloud sandbox, and Portable mode can run the whole loop on your own GPU. Conditional because cloud work is metered in credits that do not roll over, and the entry paid plan includes no monthly allocation, so heavy use means top-ups or the top plan. Right for knowledge workers who will watch the meter; we have not tested it.

Best for

A strong pick for continuous customer research where you want conversations rather than survey responses, and a poor fit for operational intake despite ranking first for that term. Conversations cost 10 credits each against 1,000 credits a month on the $99 Pro plan, so roughly 100 conversations a month before you buy more. That is research economics. If you are collecting a hundred enquiries a week, price it carefully before committing, and check that the output is the structured record your downstream system needs rather than a summary a human still has to read.

Best for

Pi is a deliberately minimal, MIT-licensed agent harness with TUI, print/JSON, RPC and SDK modes, 15+ model providers plus local servers, and built-in MCP. The deciding caveat is safety: Pi's docs say it does not ask for approval before every tool call and that tools run with your account's permissions. Right for tinkerers who build their own harness inside a VM or container; wrong for anyone wanting guardrails out of the box.

Best for

Typeform is still the best-looking, best-integrated form builder in the category and the safe default for short, branded, non-technical flows. The condition is volume. Basic costs $39 a month, or $28 a month billed annually, and includes only 100 responses per month, the same allowance as the free plan. You need Business at $129 monthly, or $91 billed annually, to reach 10,000. Long forms stay long, because the question list is fixed before anyone arrives.

Best for

Filesystem-first agent framework that turns directories into production agents — best-in-class for Vercel-native TypeScript teams, but platform coupling is real.

Best for

The easiest automation tool to start with, but per-task pricing adds up fast for high-volume workflows.

Best for

ZCode pairs an agent-first design with GLM-5.3 and GLM-5.3-Flash, on GLM Coding plans that start at $18 a month per Z.ai's docs. GLM-5.3 weights are published on Hugging Face, though under a bespoke licence to read before commercial use. Against that, Linux support is beta and reviewers have flagged credential handling for remote and messaging-triggered tasks. Conditional until security and trust catch up with the product.

Best for