AI assistant for web tasks that also works across your apps
Which AI assistants can finish a job on a real website and also use several of your tools in one request?
27 assistants ranked
Twenty-seven assistants in our catalog took both the Online task and the Connected apps tests. Muse and Instinct tie at 8.5. Neo, Ollie, Bantr and OpenInstinct follow at 8.0, then szn at 7.5 and Pally at 7.0.
Top picks
- 18.5MuseTop pick
Online tasks 8Connected apps 9
- 28.5
Online tasks 9Connected apps 8
- 38.0NeoBest free option
Online tasks 7Connected apps 9
- 48.0
Online tasks 9Connected apps 7
- 58.0
Online tasks 8Connected apps 8
- 87.0PallyFastest that's still good
Online tasks 7Connected apps 7Time to finish 7m 1s
Top pick: highest average here. Best free option: highest among assistants you can use for free. Fastest that's still good: quickest median time to finish among those averaging 7 or more.
Full ranking
| # | Assistant | Online tasks | Connected apps | Average |
|---|---|---|---|---|
| 1 | Muse | 8 | 9 | 8.5 |
| 2 | Instinct | 9 | 8 | 8.5 |
| 3 | Neo | 7 | 9 | 8.0 |
| 4 | Ollie | 9 | 7 | 8.0 |
| 5 | Bantr | 8 | 8 | 8.0 |
| 6 | OpenInstinct | 10 | 6 | 8.0 |
| 7 | szn | 7 | 8 | 7.5 |
| 8 | Pally | 7 | 7 | 7.0 |
| 9 | Tomo | 7 | 7 | 7.0 |
| 10 | Caddy | 8 | 6 | 7.0 |
| 11 | Today | 8 | 6 | 7.0 |
| 12 | OpenClaw | 7 | 7 | 7.0 |
| 13 | Hark | 8 | 5 | 6.5 |
| 14 | Buoy | 7 | 6 | 6.5 |
| 15 | Asmi | 7 | 6 | 6.5 |
| 16 | Poke | 4 | 8 | 6.0 |
| 17 | Wajo / Fo | 2 | 10 | 6.0 |
| 18 | Waldo | 6 | 6 | 6.0 |
| 19 | Del | 5 | 7 | 6.0 |
| 20 | Familiar | 7 | 4 | 5.5 |
| 21 | Gemini Spark | 6 | 5 | 5.5 |
| 22 | AVA | 3 | 7 | 5.0 |
| 23 | Joshu | 5 | 5 | 5.0 |
| 24 | Manus | 4 | 5 | 4.5 |
| 25 | Ask Ranger | 4 | 4 | 4.0 |
| 26 | Persona | 5 | 3 | 4.0 |
| 27 | Lucas | 4 | 4 | 4.0 |
How this report works
Included are assistants that took both Online task and Connected apps, ranked by the average of the two. Online task checks a real booking at a named restaurant with the right time, name and number. Connected apps checks that one request puts a to-do, a calendar block and a message in the right places. This differs from the get-things-done report: it uses only these two tests and covers every assistant that took both. Equal averages are ordered by overall score.
How to choose
- Look at both scores, not the average. OpenInstinct scores 10 on Online task but 6 on Connected apps; Wajo / Fo scores 10 on Connected apps but 2 on Online task.
- Check the channel you will use. Neo lists Telegram and iMessage, Ollie iMessage and SMS, Bantr SMS and OpenInstinct iMessage and an app.
- Check the price. OpenInstinct and Pally list free plans, Ollie lists $25 a month and Neo $49, each with a free tier.
Which AI assistant does web tasks and uses my apps?
Muse and Instinct tie at 8.5 over the two tests, then Neo, Ollie, Bantr and OpenInstinct at 8.0.
How is this different from the get-things-done report?
That report ranks four tests, including travel, purchasing and phone calls. This one uses only Online task and Connected apps.
Does a high Online task score mean it books anything?
No. It means it booked a table in our trial. Check what the vendor says about other sites.
Where is the web-browsing report?
See our report on the AI assistant that browses the web.
What users say
People have reported that when they tried to disconnect the agent from their email, they found out it was retaining a copy of their inboxes anyway.
Open sourceThe agent retrieved a one-time Luma login code from his connected Gmail inbox — without first asking him — and used it to access the account and cancel the RSVPs.
Open sourceThe agent, which Meta launched on September 8, booked a Friday date night and made a packing schedule for my upcoming move. It also emailed a friend about a trip next month and made a planning document with places we want to visit.
Open sourceFrequently asked questions
Which assistant is best for this?
Muse leads with an average of 8.5, ahead of Instinct and Neo.
Is there a good free option?
Muse is the highest-ranked assistant with a free plan, averaging 8.5.
How did you test these assistants?
Every assistant got the same published tests, word for word, scored from three checkpoints and the number of corrections it needed. Each test is on the Dimensions page.