AI Tools Like Claude — 15 Worth Trying

If you’re looking for AI like Claude, the useful question isn’t “which chatbot is best” — it’s “which tool is best at the specific job you hire Claude for.” So instead of another head-to-head ranking, this guide sorts 15 credible tools by job: writing, coding, research, and open-source models you can run yourself. For each one I’ll tell you what it does better than Claude, what it does worse, and who should actually use it.

TL;DR:

• Writing: ChatGPT, Gemini, Notion AI, Sudowrite — pick by where your drafts live.

• Coding: GitHub Copilot, Cursor, OpenAI Codex, Windsurf — pick by how agentic you want the tool.

• Research: Perplexity, NotebookLM, Elicit — pick by whether sources are the web, your files, or academic papers.

• Open-source/local: Llama, Mistral, DeepSeek, Qwen via Ollama — pick when privacy or cost beats raw capability.

How to choose an AI like Claude: match the tool to the job

Claude is a generalist — the same product handles a novel chapter, a Python refactor, and a literature review. Most tools “like” it are actually specialists wearing a chatbot interface, and that’s their advantage: a tool built around one job often beats a generalist at that job. My method for this list: I’ve used every tool below on real work, and I’ve assigned each one to the single job where it earns its spot. Prices are current as of mid-2026 but move often — verify before paying. (If you want the head-to-head ranking format instead, my 10 best Claude alternatives compared covers the direct-substitute question.)

Job	Tools	Start with
Writing & content	ChatGPT, Gemini, Notion AI, Sudowrite	ChatGPT
Coding	GitHub Copilot, Cursor, OpenAI Codex, Windsurf	Cursor
Research	Perplexity, NotebookLM, Elicit	Perplexity
Open-source / local	Llama, Mistral, DeepSeek, Qwen (via Ollama)	Llama on Ollama

Writing tools like Claude

Writing is where Claude is hardest to replace — its drafts consistently need the least editing of anything I use, which is why Claude for writing is still my default workflow. But these four earn consideration.

1. ChatGPT — the versatile second drafter

ChatGPT Plus ($20/month) is the strongest general writing substitute: solid drafting, good rewriting, plus things Claude doesn’t do at all — image generation for illustrations and a polished voice mode for dictating ideas. Its default prose runs more formulaic than Claude’s in my testing; budget an extra editing pass.

2. Google Gemini — for writing inside Docs and Gmail

If your writing lives in Google Docs and Gmail, Gemini (Google AI Pro, ~$19.99/month) edits where you type — no copy-paste loop. The integration is the product; the raw drafting quality is competitive but not Claude-level on long structured pieces.

3. Notion AI — for notes and team docs

Notion AI is an add-on inside the Notion workspace: summarize meeting notes, turn bullets into prose, keep a team wiki consistent. It’s not a frontier model experience and won’t replace Claude for serious drafting, but if your team already lives in Notion, it removes the friction of leaving.

4. Sudowrite — for fiction specifically

Sudowrite is built exclusively for fiction — scene expansion, character voice consistency, “show don’t tell” rewrites. Claude is a strong fiction collaborator, but Sudowrite’s purpose-built tooling (story bibles, chapter-level continuity) is something a chat interface can’t match. Niche, but the clear winner in that niche.

Coding tools like Claude

Fair warning: several of these run Claude models under the hood, so “like Claude” here often means “Claude in a different shape.” The real competition is between form factors — autocomplete, AI-native editor, or autonomous agent. The full benchmark is in best AI for coding in 2026.

5. GitHub Copilot — the workflow staple

Copilot Pro ($10/month, free for verified students) is the cheapest serious coding assistant and lives natively in VS Code and GitHub. It has grown from autocomplete into a multi-model assistant — you can select Claude models inside it. Less autonomous than Claude Code for multi-file work, but the price and ubiquity are hard to argue with. My full Claude vs Copilot comparison weighs the two head-to-head for developers.

6. Cursor — the AI-native editor

Cursor ($20/month Pro) is a VS Code fork rebuilt around AI: codebase-aware chat, multi-file edits, an agent mode. Many Cursor users run Claude models inside it, which makes the real question Cursor’s editor experience versus Claude Code’s terminal autonomy rather than model quality.

7. OpenAI Codex — the direct Claude Code rival

Codex is OpenAI’s agentic coding tool — same category as Claude Code: give it a task, it works across your repo, runs tests, and proposes changes. Capability leadership swings with each model release; in my recent testing Claude Code still handles long multi-step refactors more reliably, but Codex is a genuine peer, not a copy.

8. Windsurf — the accessible agentic editor

Windsurf competes with Cursor as an AI-native editor with an agent (“Cascade”) that plans and executes multi-file changes. Its pricing has historically undercut Cursor’s, and its flow feels friendlier for developers new to agentic tools. Worth a trial if Cursor’s pace overwhelms.

Research tools like Claude

9. Perplexity — web research with receipts

Perplexity ($20/month Pro, useful free tier) answers questions with numbered citations from live web sources. Claude’s Research mode has closed much of the gap, but Perplexity remains the tool I reach for when verifiability matters more than synthesis. Full comparison in Claude vs Perplexity.

10. NotebookLM — research grounded in your own sources

Google’s NotebookLM answers only from documents you upload, with citations back to the exact passage — which nearly eliminates hallucinated claims. Its audio overview feature (turning your sources into a podcast-style discussion) is genuinely novel. Claude’s Projects cover similar ground with more reasoning power; NotebookLM wins on source discipline and price (free core product).

11. Elicit — for academic literature

Elicit searches academic papers and extracts structured data from them — populations, methods, outcomes — into comparison tables. For a literature review, it does in an afternoon what used to take a week. It’s a specialist: you’d still hand the synthesis and writing to Claude.

Open-source and local models like Claude

The open-weight world gives up some capability for total control: no subscription, no data leaving your machine, no usage caps. The easiest on-ramp is Ollama — install it, pull a model, and you’re chatting locally in minutes; LM Studio does the same with a friendlier GUI. A 16GB+ RAM machine runs mid-size models comfortably, and a machine with a decent GPU or Apple Silicon runs larger ones at usable speeds. Set expectations honestly: local models are where cloud frontier models were a generation or two ago — excellent for drafting, summarizing, and routine coding, thinner on hard multi-step reasoning.

12. Llama — the open-source default

Meta’s Llama family is the most widely supported open-weight line: every local tool, fine-tuning guide, and hosting provider supports it first. Smaller variants run on laptops; larger ones approach older frontier quality. If you’re going local, start here.

13. Mistral — open weights with a European home

Mistral publishes strong open-weight models alongside its hosted Le Chat product, and its permissively licensed small models punch above their size for coding and summarization. The EU jurisdiction is a bonus for compliance-sensitive work.

14. DeepSeek — frontier-adjacent reasoning, open and cheap

DeepSeek’s open reasoning models shocked everyone by matching benchmark scores of far more expensive systems. Run locally, the jurisdiction concerns of its hosted app disappear — the weights are just weights. Heavier to run well than Llama, but the strongest open reasoning per gigabyte as of mid-2026.

15. Qwen — the multilingual open family

Alibaba’s Qwen models are the sleeper pick: consistently near the top of open-model leaderboards, with excellent multilingual coverage and strong small coding models. If Llama underwhelms on your language or task, Qwen is the next pull.

How to combine these tools without wasting money

The stack that works for most people is one paid generalist plus free specialists. In practice: keep Claude Pro or ChatGPT Plus as the daily driver, then bolt on NotebookLM (free) for source-grounded research, GitHub Copilot’s student tier if you qualify, and a local Llama install for anything you’d rather not send to a cloud at all. Total cost: one $20 subscription. The inverse pattern — three paid subscriptions, each used twice a week — is the most common waste I see, and cancelling down to one plus free tiers rarely hurts anyone’s actual output. Review the stack quarterly: this market changes fast enough that a tool you dismissed in January can be the category leader by June.

Common mistakes when picking a Claude-like AI

• Hiring a generalist for a specialist’s job. If 90% of your usage is literature review, Elicit plus a free chatbot beats any single $20 subscription.
• Not realizing the tool runs Claude anyway. Cursor, Copilot, and Perplexity can all route to Claude models. Sometimes “an AI like Claude” is literally Claude with a different bill.
• Overestimating local models. A laptop-size Llama is not Fable 5. Test your hardest real task locally before cancelling anything.
• Stacking subscriptions. Three $20 tools with 80% overlap is the most common failure mode I see. One generalist plus one specialist covers almost everyone.
• Skipping the free evaluation. Every tool here has a free tier or trial. A structured week of testing (template in my Claude free trial guide — it works for any tool) beats a hundred review articles.

Next steps

Pick your job, trial the “start with” tool from the table for a week, and keep Claude for whatever remains — for me that’s long-form writing and agentic coding, where nothing on this list fully substitutes. For more on trials, student pricing, and whether you can invest in Anthropic itself, browse the full alternatives section.

FAQ