State of the mod ecosystem (October 2026)

As of: community catalog scan of 2026-10-04, Claude Code 2.1.289

The numbers in this article come from recounting data/mods.json (CC0) in the awesome-claude-code-mods catalog with Python. The catalog searches GitHub for mods, runs claude plugin validate on each repository, and extracts the hooks and $ calls from that output. The file was generated at 2026-10-04 08:26 (UTC). Mods created after that are not included.

What we counted

The catalog has 1,919 entries found in 1,171 repositories. Not all of them are mods you would install.

Kind	Entries	Meaning
mod	1,752	Regular mods
fixture	93	Test examples in folders like tests/ and examples/
catalog	39	Entries from repositories that re-bundle other people’s mods
duplicate	28	The same mod as another entry
builtin	4	Built-in mods in the anthropics/claude-code repository
mirror	3	Copies of built-in mods

This article counts only entries that are mod or builtin, not archived, not duplicates, and that passed validation without warnings. What remains is 1,488 mods, 777 repositories, and 690 authors. Unless stated otherwise, the numbers below are based on these 1,488 mods. The mod directory on this site also lists the 224 mods that passed with warnings, so it shows 1,712.

Three days of flooding

The catalog’s createdAt is when the repository was created. It is not always the same as the day the mod was uploaded. For example, the 102 repositories created before September 2026 added their mods later.

Repository creation date (UTC)	Repositories	Mods in those repositories
Before September 2026	102	154
September 2026	108	190
October 1	36	52
October 2	258	453
October 3	228	568
October 4 (until 08:26)	42	68
October 5	3	3

486 repositories were created on October 2 and 3 alone, which is 63% of all 777. The October 3 figure includes 117 mods from a single repository created that day, mako-code/gobworks. KilimcininKorOglu/claude-code-mods, created on September 18, also added 53 mods in a single day.

Claude Code 2.1.287 was published to the npm registry at 2026-10-01 16:59 (UTC). As the FAQ explains, that is the version where function hooks are on by default. Of the 36 repositories created on October 1, 29 were created after 17:00. Repositories created after that time number 560 of 777 (72%), and the mods in them are 1,137 of 1,488 (76%). The timing overlaps, but the catalog does not record why a repository was created, so we cannot name a cause.

One mod per repository is typical

Mods per repository	Repositories	Mods	Share of mods
1	626	626	42.1%
2-9	133	429	28.8%
10-17	15	193	13.0%
50 or more	3	240	16.1%

81% of repositories hold a single mod. At the same time, the top 3 repositories hold 240 mods.

Repository	Mods	Character
mako-code/gobworks	117	A collection split into groups such as jev, abilities, ambient, and autopilot under mods/
whyashthakker/awesome-claude-code-mods	70	A collection of small tools, one mod per folder
KilimcininKorOglu/claude-code-mods	53	A collection with gemini-* integrations and shared parts such as sidebar split out

Mods in large collection repositories sometimes share parts. 36 mods in the KilimcininKorOglu repository call $.sidebar, which the repository’s own sidebar mod adds to $. If you install only one of them, that part may be missing, so check the dependencies in the README first.

Stars are per repository

The catalog’s stars is the repository’s star count. A repository with 117 mods gives all 117 the same number. That is why we counted by the 777 repositories.

Repository stars	Repositories	Share
0	538	69.2%
1-9	188	24.2%
10-99	36	4.6%
100-999	9	1.2%
1,000 or more	6	0.8%

Of the 6 repositories with 1,000 or more stars, 5 are large projects that do much more than mods: anthropics/claude-code (149,303), ruvnet/RuView (96,198), ruvnet/ruflo (73,820), anthropics/claude-plugins-official (37,363), and kunchenguid/firstmate (7,488). Among repositories where a mod is the main attraction, zenbu-labs/terminal-browser (3,626) is the largest. Conversely, gobworks, with the most mods, had 0 stars at scan time. Most of these repositories were created within three days, so stars had no time to accumulate. Stars do not tell you the quality of a mod.

One in five repositories has no license

The catalog records the license (SPDX ID) that GitHub detected.

License	Repositories	Share
MIT	555	71.4%
None	164	21.1%
Apache-2.0	25	3.2%
NOASSERTION	24	3.1%
AGPL-3.0	3	0.4%
GPL-3.0 and Unlicense	2 each	0.3%
GPL-2.0 and MIT-0	1 each	0.1%

Counted by mod, 324 mods (21.8%) have no license. A repository without a license file falls under the default of copyright law: unless the author grants permission, others have no right to copy the code or redistribute a modified version. Under GitHub’s Terms of Service you may view and fork a public repository. Even so, you must not paste that code into your own mod or distribute a modified version. If you need it, open an issue asking the author to add a license. NOASSERTION means a license file exists but GitHub could not identify its type. Open the file and check it yourself.

Where mods draw

1,129 mods (75.9%) hook ui.render. The other 359 use only toasts, status lines, and log lines, or show nothing on screen.

Draw location	Mods	What it is
Pane	673	A window that opens to the side or below
AbovePrompt	600	A band above the input box
Spinner	76	The working indicator
PromptHint	75	The hint line below the input box
SessionMode	65	The mode indicator at the bottom
Transcript rows	153	AssistantMessage 67, ToolUse 59, UserMessage 50, ToolResult 40, ToolGroup 37, CommandOutput 33, TurnDuration 32, InfoNotice 7
AskUserQuestion	11	The question dialog
ToolProgress	7	The tool progress indicator

363 mods draw only Pane, 274 draw only AbovePrompt, and 266 draw both. A common layout is a pane that shows details with a one-line summary in the band. The 153 mods that redraw transcript rows change how Claude’s replies or tool results look on screen. The permission prompt is not on the list. The type file also states that only the engine draws that dialog.

Frequently hooked events

The catalog shows 114 event types. A mod hooks a median of 5 events and calls a median of 11 $ methods.

Event	Mods	Share	Typical use
session.start	1,284	86.3%	Register commands and tools, start timers
ui.render	1,129	75.9%	Draw on screen
command.run	1,078	72.4%	Handle slash commands
tool.call	786	52.8%	Inspect, block, or change tool calls
turn.complete	738	49.6%	Tally and notify after a turn ends
prompt.submit	474	31.9%	Inspect or change the submitted prompt
turn.start	367	24.7%	Show that a turn has started
session.end	279	18.8%	Clean up and save
turn.step	204	13.7%	Model request stream
ui.close	182	12.2%	Handle a window closing
session.measure	180	12.1%	Receive changes in usage and limit figures
session.compact	144	9.7%	When and how compaction happens
agent.spawn	109	7.3%	Choose or block the subagent model
classic.SessionStart	103	6.9%	Receive a settings hook event as a function

196 mods hook at least one classic.* event. A function hook receives, as is, the input that a settings-file hook (such as Stop or PermissionRequest) would receive.

Note

The catalog numbers come from reading source. They are not the result of running the mods. These numbers cannot tell you how often a hook actually runs or whether the screen renders correctly.

How to use this when choosing

• Star counts reflect the repository’s name value. To choose a single mod, look at the last commit in that mod’s folder and its source.
• For a mod in a large collection repository, check whether it depends on other mods in the same repository.
• You can install and try a mod without a license, but you must not take its code and reuse it.
• Access scope continues in What mods actually touch, and the causes of validation failures in Mods that fail validation.