AI coding tools: 422 of them, and how to choose

Coding is the largest market in the catalog — 422 live products, more than any other category — and one of the most contested, at 60 opportunity against 44 saturation. It is also, unusually, one of the safest. Very few AI coding tools die. This is what 422 products have divided the job into, and how to tell which part of it a given tool is actually competing for.

There are four products here, not one

Treating “AI coding tools” as a single market is the main reason comparisons go wrong. The 422 products split into four jobs that barely compete with each other.

Completion inside the editor — the incumbent category, now effectively commodity. Agentic editors, where the tool holds the whole repository and makes multi-file changes on instruction. App builders, which take a prompt and return a running application, aimed as much at people who do not write code as at people who do. And autonomous engineers that take a ticket and open a pull request.

The top of the market by opportunity score is dominated by the last three. Completion is missing from the leaders not because it stopped being useful, but because it stopped being differentiated — it is the part of the market where a model provider can ship the feature natively, and largely has.

Crowded, yet hardly anyone dies

Coding has the most competitors and the seventh-lowest failure rate of the 29 markets we track. That combination is rare enough to be worth explaining. The catalog-wide pattern is that products selling a generated artefact to a consumer die fastest; developer tools do the opposite. The buyer is technical, evaluates quickly, and adopts individually rather than through procurement. Distribution is cheap — a plugin, a CLI, a GitHub integration — so a small team can reach real users without a sales organisation.

The consequence is that competition here shows up as margin pressure rather than as shutdowns. Products do not disappear, they get absorbed into someone’s free tier. That is a different risk to manage, and a slower one.

Almost everything is free to start

375 of 418 offer a free tier — the second-highest rate in the catalog after Music. Only 41 charge from the first minute. Developers will not evaluate a tool they cannot run on their own repository first, so a free tier here is not a growth tactic, it is the cost of being considered at all.

For anyone entering, that sets the floor: your free tier has to be genuinely useful, and it has to absorb inference cost from day one. The market’s economics are set by whoever is most willing to lose money on evaluation.

Growth-stage and bootstrapped, in roughly equal measure

113 are at growth stage and 96 are bootstrapped. That second number is the interesting one: 22.7% of this market never raised, which is possible because a developer tool can find its first thousand users without spending anything on reaching them. In Healthcare the bootstrapped count is zero, and in Real Estate and Human Resources it is one apiece.

How to choose between them

Match the tool to the job rather than to the benchmark. If the work is editing a codebase you already know, an agentic editor earns its price in the first week. If the work is getting from nothing to something demonstrable, an app builder will beat it and the code quality question matters less than people claim. If the work is a queue of small, well-specified tickets, an autonomous engineer is worth trialling — and if the tickets are not well specified, it is not.

One practical warning from the data: 54% of coding products ship an API. In a market where products get absorbed rather than shut down, an API is what lets you keep your workflow when the tool underneath changes hands or changes pricing. Our API report has the cross-market picture.

The market, product by product

Ranked by opportunity score. Each name links to the product’s own site.

Photo: Christina Morillo / Pexels. Colour-graded for Falcoscan.

← All articles