# GitHub Star Velocity: 23 Open-Source AI Agents Measured 17 Days Apart

> Source: The Agents Index — https://theagentsindex.com/blog/github-star-velocity-open-source-ai-agents-2026 (sourced research)
> By The Agents Index Editorial · Published 2026-07-31 · Updated 2026-08-17

_Stars for all 30 open-source AI agents in our index, counted 2026-08-01 and 2026-08-18. Three gained 1,076 stars in 17 days without shipping a commit._

Of the 89 tools published in our index, 89 have a determined open-source status and **30 of those are open source**. On 2026-08-01 we queried the GitHub REST API for the star count, fork count and last-push timestamp of every one we could then identify. On 2026-08-18 we did it again, the same way, from the same API. This page is the second measurement set against the first.

Two dated counts is the whole point. A single star count is a number anyone can fetch in a second, and it tells you almost nothing on its own. **The rate of change needs two measurements taken by the same method at two known dates**, and for 23 of these tools the gap is exactly 17 days: 2026-08-01 to 2026-08-18.

## Three projects gained 1,076 stars while shipping nothing

The most decision-relevant finding here is not who grew fastest. It is the three tools whose default branch received no push at all during the 17 days between our two counts, and which gained stars anyway:

| Tool | Last push to default branch | Days quiet on 2026-08-18 | Stars gained over the 17 days |
|---|---|---:|---:|
| [Microsoft AutoGen](/microsoft-autogen) | 2026-04-15 | 125 | +349 |
| [Aider](/aider) | 2026-05-22 | 88 | +459 |
| [GPT Researcher](/gpt-researcher) | 2026-07-18 | 31 | +268 |

Between them that is **1,076 stars for zero commits**. Stars measure discovery, so a project can keep accumulating them long after the people who wrote it have stopped. A buyer reading only the star count on any of these three rows would see a number going up.

The three cases are not equally strong, and it matters which one you are looking at. AutoGen at 125 days is consistent with the Microsoft-declared maintenance mode already documented on its listing. Aider at 88 days is a genuine gap with no announced explanation we found. GPT Researcher at 31 days is barely a gap at all: plenty of healthy projects go a month between pushes, and we include it because it met the same test as the other two, not because a month is alarming.

A fourth repository in the wider set was also quiet and is the useful counter-example. [Cartesia](/cartesia)'s Line SDK has not been pushed since 2026-07-20, 29 days, and it gained exactly zero stars over the same period. Quiet does not imply growing. It is the combination that is worth knowing about.

One caveat we would rather state than have someone find. `pushed_at` reports the default branch only. A project can be actively maintained in forks, on release branches or by a maintainer who batches merges, and none of that shows up here.

## The 17-day leaderboard

Stars gained per day for the 23 tools we counted on both 2026-08-01 and 2026-08-18. Every baseline in this table is our own published figure from the first count, not a number lifted from anywhere else.

| Tool | Stars 2026-08-01 | Stars 2026-08-18 | Gained | Stars/day |
|---|---:|---:|---:|---:|
| [OpenCode](/opencode) | 191,820 | 198,731 | +6,911 | 406.5 |
| [OpenAI Codex CLI](/codex-cli) | 103,012 | 106,647 | +3,635 | 213.8 |
| [n8n](/n8n) | 198,915 | 201,077 | +2,162 | 127.2 |
| [Dify](/dify) | 150,967 | 152,807 | +1,840 | 108.2 |
| [OpenHands](/openhands) | 82,753 | 84,399 | +1,646 | 96.8 |
| [LiveKit Agents](/livekit-agents) | 11,586 | 13,063 | +1,477 | 86.9 |
| [LangChain](/langchain) | 143,147 | 144,478 | +1,331 | 78.3 |
| [Cline](/cline) | 65,368 | 66,406 | +1,038 | 61.1 |
| [Zed](/zed) | 87,860 | 88,823 | +963 | 56.6 |
| [CrewAI](/crewai) | 56,450 | 57,255 | +805 | 47.4 |
| [Mastra](/mastra) | 26,804 | 27,272 | +468 | 27.5 |
| [Warp](/warp) | 63,844 | 64,310 | +466 | 27.4 |
| [LlamaIndex](/llamaindex) | 51,267 | 51,727 | +460 | 27.1 |
| [Aider](/aider) | 47,848 | 48,307 | +459 | 27.0 |
| [Pydantic AI](/pydantic-ai) | 18,939 | 19,369 | +430 | 25.3 |
| [OpenAI Agents SDK](/openai-agents-sdk) | 28,325 | 28,746 | +421 | 24.8 |
| [Microsoft AutoGen](/microsoft-autogen) | 60,142 | 60,491 | +349 | 20.5 |
| [Microsoft Agent Framework](/microsoft-agent-framework) | 12,527 | 12,869 | +342 | 20.1 |
| [GPT Researcher](/gpt-researcher) | 28,756 | 29,024 | +268 | 15.8 |
| [Letta](/letta) | 24,046 | 24,293 | +247 | 14.5 |
| [Google ADK](/google-adk) | 20,962 | 21,169 | +207 | 12.2 |
| [Juggler](/juggler) | 544 | 567 | +23 | 1.4 |
| [Zot](/zot) | 299 | 318 | +19 | 1.1 |

**OpenCode gained 6,911 stars in 17 days, nearly twice the next tool on the list.** At 406.5 stars a day it is adding roughly the whole of Zot's lifetime star count every nineteen hours. The two fastest projects here are both terminal-native coding agents, and the third and fourth are both workflow platforms with far longer histories.

Our previous count put OpenCode at about 237 stars a day. It has accelerated since.

## Fastest growth is a different question from most growth

Rank the same 23 by percentage gained instead of absolute stars and the order changes at the top. **LiveKit Agents grew 12.75% in 17 days, the fastest relative growth in the set**, against OpenCode's 3.60% and LangChain's 0.93%.

That difference is the useful one for anything early. A project adding 1,477 stars off a base of 11,586 is being discovered right now. A project adding 1,331 off a base of 143,147 has already been discovered, and its absolute number mostly reflects how long it has been around.

## Four tools have no prior figure, so they get no rate

Four open-source listings in the index have never had a star count published by us, so this page gives their live snapshot and no growth rate. Inventing a rate from a single measurement is the one thing this method cannot do.

| Tool | Stars 2026-08-18 | Forks | Last push |
|---|---:|---:|---|
| [OpenClaw](/openclaw) | 386,636 | 81,244 | 2026-08-18 |
| [Goose](/goose) | 52,954 | 6,027 | 2026-08-18 |
| [LangGraph](/langgraph) | 39,937 | 6,725 | 2026-08-16 |
| [AgentsKit](/agentskit) | 24 | 1 | 2026-08-15 |

**OpenClaw's repository holds 386,636 stars, which makes it the largest repository in our index by a factor of nearly two over n8n.** It was created on 2025-11-24, so it reached that count in under nine months. That figure surprised us enough that we checked it against a second, independent source before publishing it, and it held.

Goose is the newest of the four and the largest first-time entry: 52,954 stars puts it thirteenth by size across all 30 repositories, above Aider and below CrewAI. It joined the index on 2026-08-17, after our first count, so its rate has to wait for the next edition.

Three more carry a dated baseline from a different day, because they joined the index after the 2026-08-01 count: Botpress (14,841 on 2026-08-02, now 14,872), Agno (41,596 on 2026-08-06, now 41,763) and Cartesia's Line SDK (103 on 2026-08-05, unchanged at 103).

## Every listing in the index now has a determined open-source status

Twenty-six of the 30 repositories pushed a commit within three days of this count. The quiet ones are the exception here, not the pattern.

The open-source denominator on this page changed since the previous edition, and it changed in the direction that makes the number stronger. **All 89 published listings now carry a researched open-source status, so the denominator for the open-source share is the whole index rather than a subset of it.** The previous edition had to exclude six listings whose status we had not established: Augment Code, Browserbase, Gemini Code Assist, Julius AI, SalesTouch and TinyFish. All six have since been researched and all six are proprietary, which moves the count of determined listings from 82 to 89 and leaves no unknown bucket at all.

Two of those six are worth naming precisely, because each ships an open-source artifact that does not extend to the product itself. Browserbase publishes Stagehand, an automation framework anyone can use without a Browserbase account, while the hosted browser infrastructure it sells stays closed. Gemini Code Assist bundles access to Gemini CLI, a separately branded Apache-2.0 Google project, while Code Assist itself ships nothing open. Neither has a repository this page can track.

For license terms rather than activity, our [open-source license census](/blog/open-source-license-census-ai-agents-2026/) is the reference. It is worth reading alongside this page, because two of the largest repositories here carry a licensing caveat a star count will not show you: n8n's fair-code license and Dify's modified Apache terms are both source-available rather than OSI-approved.

## What this measures, and what it doesn't

A star is a free, one-click action. It measures interest, not code quality, security posture, test coverage or whether your team should build on the thing. We publish velocity because it is a genuinely different signal from a raw count, not because it ranks anything.

The one thing it does well is catch the divergence in the first table. When a project's stars climb while its commits stop, the star count is describing its past and the commit log is describing its present. Use this page to see which of those two you are looking at, then use each tool's own listing for pricing, licensing and the sourced verdict before you decide anything.

## Method

Every figure marked 2026-08-18 was read from the [GitHub REST API](https://docs.github.com/en/rest/repos/repos#get-a-repository) (`api.github.com/repos/{owner}/{repo}`) at 15:08 UTC on 2026-08-18: `stargazers_count`, `forks_count` and `pushed_at`, which we report as "last push". All 30 repositories were read in a single pass, so no figure on this page is older than any other. Every figure marked 2026-08-01 is our own previously published count, read the same way on that date. Rates are the difference divided by the exact number of days between the two dates, which is 17 for the 23 tools in the leaderboard.

Repository identity came from each listing's own sourced research where we hold it, which covers 25 of the 30. The remaining five we resolved by matching the repository's declared homepage against the website URL on our listing, then confirming the star count was continuous with our 2026-08-01 figure where we had one. Two repositories deserve naming: OpenCode is `anomalyco/opencode`, not the archived `opencode-ai/opencode` that shares its name, and OpenHands resolves through a permanent redirect from `All-Hands-AI/OpenHands` to `OpenHands/OpenHands` after an organization rename. Where a tool ships several language SDKs in separate repositories, as Google ADK does, we count the Python repository only.

Two counts were verified twice because they were surprising. OpenClaw returned 386,636 from the GitHub API and 387k from shields.io, an independent badge service that queries GitHub separately. Goose returned 52,954 and 53k from the same pair. Both agree.

This is a dated measurement, not a running index. The star counts on this page were true at 15:08 UTC on 2026-08-18 and will be wrong tomorrow, which is why every one of them carries its date in the same sentence. The corpus figures (89 published, 89 determined, 30 open source) are checked against our live database daily; the GitHub figures are not, because they come from outside it. A future edition will need its own fresh pair of dates.

---

Tags: ai-agents, open-source, github, data-report, market-analysis