Apps · 2026-09-02
The apps burning the tokens.
The AI apps routing the most LLM tokens through OpenRouter, and this week's breakouts. Data as of 2026-09-02.
Popular apps
| # | App | Category | Tokens (window) | Requests | Avg tok/req | 7d |
|---|---|---|---|---|---|---|
| 1 | Hermes Agent | coding | 49.23T | 492.59M | 99.9K | -26.4% |
| 2 | Claude Code | coding | 13.95T | 138.02M | 101.1K | -35.6% |
| 3 | Kilo Code | coding | 9.25T | 113.94M | 81.2K | -16.7% |
| 4 | Zazen (Freebuff fork) | — | 8.25T | 68.77M | 120.0K | -5.7% |
| 5 | Cline | coding | 7.52T | 70.58M | 106.5K | +28.4% |
| 6 | pi | coding | 5.45T | 67.33M | 80.9K | -16.9% |
| 7 | OpenClaw | coding | 4.59T | 75.44M | 60.8K | -7.4% |
| 8 | omp | coding | 4.28T | 30.51M | 140.2K | -46.2% |
| 9 | DeepSeek Harness | — | 3.50T | 26.70M | 131.2K | -73.9% |
| 10 | Framer | — | 2.57T | 19.22M | 133.6K | -5.7% |
| 11 | Codex | coding | 2.46T | 36.28M | 67.8K | -5.6% |
| 12 | Nous Research API | productivity | 1.81T | 39.77M | 45.5K | -70.5% |
| 13 | ZCode | coding | 1.29T | 7.49M | 171.9K | -78.0% |
| 14 | ISEKAI ZERO | entertainment | 1.26T | 39.07M | 32.3K | +12.7% |
| 15 | Hello Minds, powered by Ethoswarm | creative | 1.23T | 5.64M | 218.2K | -15.8% |
| 16 | Descript | creative | 1.23T | 19.15M | 64.2K | +0.1% |
| 17 | Lemonade | coding | 1.07T | 28.81M | 37.2K | +112.2% |
| 18 | Janitor AI | entertainment | 1.01T | 66.46M | 15.3K | -1.3% |
| 19 | schema-markup-generation | — | 924.6B | 15.99M | 57.8K | -86.2% |
| 20 | Cursor | coding | 915.9B | 6.93M | 132.2K | -7.5% |
| 21 | HighLevel | — | 867.2B | 17.01M | 51.0K | +13.9% |
| 22 | Portkey AI | coding | 797.3B | 21.34M | 37.4K | +397.5% |
| 23 | Command Code | coding | 715.5B | 6.67M | 107.3K | -74.4% |
| 24 | OpenHands | coding | 715.4B | 9.80M | 73.0K | +5.3% |
| 25 | Sahasra | — | 673.0B | 11.74M | 57.4K | -26.2% |
| 26 | Zed Editor | coding | 580.9B | 4.20M | 138.2K | -21.9% |
| 27 | Peezy Gateway | coding | 571.6B | 10.59M | 54.0K | -6.5% |
| 28 | HackerAI | — | 570.4B | 13.29M | 42.9K | -26.3% |
| 29 | LangChain | — | 465.1B | 49.16M | 9.5K | -2.0% |
| 30 | Craft | creative | 463.1B | 31.79M | 14.6K | +2.3% |
| 31 | Ito | coding | 448.1B | 12.29M | 36.5K | -8.5% |
| 32 | Mira is the leading AI agent inside Telegram | creative | 434.8B | 28.28M | 15.4K | +21.6% |
| 33 | Zoo Code | — | 427.4B | 4.05M | 105.6K | -13.9% |
| 34 | OceanAPI | — | 383.4B | 2.93M | 130.9K | -61.2% |
| 35 | GDevelop | coding | 352.3B | 6.93M | 50.8K | -3.5% |
| 36 | SillyTavern | entertainment | 347.6B | 18.89M | 18.4K | -0.9% |
| 37 | Strix | coding | 346.9B | 3.34M | 103.9K | -28.6% |
| 38 | Cheaper Inference | — | 314.8B | 2.57M | 122.7K | -76.3% |
| 39 | claude-mem | — | 309.5B | 4.81M | 64.3K | +58.1% |
| 40 | Studs.gg | coding | 287.1B | 8.52M | 33.7K | +44.9% |
| 41 | websim | — | 281.2B | 11.32M | 24.9K | +14.5% |
| 42 | Open WebUI | productivity | 268.0B | 6.87M | 39.0K | +2.2% |
| 43 | Roo Code | coding | 212.9B | 1.51M | 141.4K | -27.1% |
| 44 | HammerAI | entertainment | 177.7B | 49.65M | 3.6K | -28.2% |
| 45 | shapes inc | entertainment | 176.5B | 29.93M | 5.9K | +26.6% |
| 46 | Bary | entertainment | 175.0B | 8.03M | 21.8K | +5.2% |
| 47 | Halluna | productivity | 172.1B | 15.51M | 11.1K | +20.8% |
| 48 | Pieces.app: Ambient Artificial Memory | productivity | 164.1B | 18.90M | 8.7K | -7.8% |
| 49 | goose | coding | 163.8B | 2.38M | 68.8K | -0.7% |
| 50 | extra.email | productivity | 159.1B | 22.99M | 6.9K | +31.9% |
| 51 | Codebuff | coding | 158.6B | 2.37M | 66.8K | +125.7% |
| 52 | MavenBio | — | 140.9B | 3.44M | 41.0K | -12.6% |
| 53 | Qwen Code | coding | 140.3B | 1.15M | 121.7K | -17.5% |
| 54 | Pocket AI | — | 136.4B | 13.82M | 9.9K | +159.8% |
| 55 | LOVE | — | 129.8B | 5.30M | 24.5K | +15.1% |
| 56 | mimocode | coding | 128.5B | 890.4K | 144.4K | -41.6% |
| 57 | Chub AI | entertainment | 127.1B | 9.53M | 13.3K | +0.7% |
| 58 | ChainPatrol | — | 126.5B | 33.34M | 3.8K | +10.9% |
| 59 | ppq.ai | — | 122.7B | 3.13M | 39.2K | +58.2% |
| 60 | Letaido | coding | 119.2B | 1.58M | 75.5K | -10.0% |
| 61 | Landingsite | — | 116.9B | 1.16M | 100.5K | +51.2% |
| 62 | HeyDitto | — | 116.0B | 26.95M | 4.3K | -40.0% |
| 63 | Sekai | — | 114.7B | 13.48M | 8.5K | -59.4% |
| 64 | Strate Systems / Codex CLI | — | 103.4B | 478.3K | 216.2K | +16.3% |
| 65 | jcode | — | 98.4B | 764.4K | 128.7K | -46.1% |
| 66 | Crush | coding | 95.6B | 874.5K | 109.3K | -60.2% |
| 67 | OrationAI | — | 92.7B | 2.69M | 34.5K | -11.5% |
| 68 | Avian.io | — | 91.0B | 3.58M | 25.4K | -59.5% |
| 69 | My Dream Companion | — | 90.1B | 13.76M | 6.6K | -1.4% |
| 70 | unit-reason-graph | — | 88.9B | 9.16M | 9.7K | +2.1% |
| 71 | Hermes Web UI | coding | 86.6B | 589.1K | 147.0K | -71.9% |
| 72 | ProVision.ai | — | 82.5B | 13.02M | 6.3K | +106.3% |
| 73 | OP.GG | productivity | 82.3B | 5.37M | 15.3K | -8.4% |
| 74 | Miniapps.ai | — | 80.8B | 12.48M | 6.5K | +3.2% |
| 75 | LUNATALK | — | 80.8B | 4.59M | 17.6K | +1.2% |
| 76 | KoalaBear | — | 79.9B | 16.43M | 4.9K | +6.5% |
| 77 | OpenCode | — | 79.7B | 419.3K | 190.0K | -0.2% |
| 78 | spacedb-search | — | 74.9B | 6.05M | 12.4K | -50.0% |
| 79 | Cherry Studio | — | 74.1B | 1.39M | 53.3K | -35.2% |
| 80 | grok-openrouter-ox | — | 73.9B | 719.5K | 102.7K | -65.3% |
| 81 | Agent Zero | coding | 72.2B | 2.00M | 36.1K | +9.5% |
| 82 | MarketingSecrets AI | — | 71.3B | 2.03M | 35.2K | -22.0% |
| 83 | Orchid | — | 70.3B | 9.12M | 7.7K | +9.9% |
| 84 | payready-document-extraction | — | 69.9B | 10.58M | 6.6K | -31.1% |
| 85 | Sentry MCP | — | 69.7B | 12.63M | 5.5K | +12.2% |
| 86 | SquadStack AI | — | 69.4B | 10.16M | 6.8K | +84.0% |
| 87 | SkyrimNet | entertainment | 69.0B | 16.31M | 4.2K | +24.4% |
| 88 | Multi-Agent Arena - Olam Labs | coding | 67.8B | 8.38M | 8.1K | +32.7% |
| 89 | Tracer III RLVR Pipeline | — | 65.4B | 6.08M | 10.8K | -70.6% |
| 90 | JobLeads LLM | productivity | 65.2B | 7.63M | 8.5K | +114.7% |
| 91 | Buzz POC check | — | 65.2B | 994.0K | 65.6K | +41.6% |
| 92 | LawSathi Judgements Extraction | — | 64.3B | 4.18M | 15.4K | -58.6% |
| 93 | Shelf Insights | — | 64.1B | 5.72M | 11.2K | -20.2% |
| 94 | Continue | — | 63.4B | 1.22M | 52.2K | +76.2% |
| 95 | nanobot | productivity | 62.0B | 1.59M | 39.0K | +55.8% |
| 96 | Traycer | coding | 61.5B | 362.0K | 169.8K | +70.9% |
| 97 | Chatbox AI | productivity | 60.9B | 1.45M | 42.0K | -35.2% |
| 98 | BuddyPro AI | productivity | 59.2B | 4.75M | 12.5K | +0.9% |
| 99 | Omelas AI | — | 59.2B | 18.93M | 3.1K | +130.3% |
| 100 | https://muah.ai | — | 58.9B | 10.40M | 5.7K | -4.9% |
as of 2026-09-02
Heavy agents vs light chat
Every app placed by average tokens per request (x) and total window tokens (y), both log-scale.
Hello Minds, powered by Ethoswarm runs the heaviest average request at ~218k tokens/call.
How to read: Top-right = heavy agent workloads. Top-left = mass chatbot traffic. Bottom-right = niche high-context users. Bottom-left = light utilities.
What a session actually costs in Hermes Agent
Median cost of one coding session by session length. 30-day window, log scale.
Agent
GPT-5.6 Sol Pro costs 4246× more per session than gpt-oss-20b in this agent.
How to read: Each row is a model. Dots run from short sessions to long ones — the further right, the more a median session costs. The number on the right is the 10–49 turns median.
1 turn2–9 turns10–49 turns50+ turns
ModelMedian
- $5.87
- $4.30
- $2.18
- $1.68
- $1.67
- $1.21
- $1.19
- $1.01
- $0.981
- $0.829
