Strawberry 1.0 AI Browser for Windows: Agent Features, Pricing and IT Security Review

What launched and who it's for​

Strawberry has launched 1.0 of its AI agent platform for non-technical teams such as sales, recruiting, customer support and marketing. The company calls its agents "companions." The Next Web says users train a companion to handle a specific job and that it learns how a team works over time.

The headline idea is a companion trained once and shared across a team. CEO Charles Maddock put it this way: the person who is great with Claude keeps all of it on their own laptop, whereas in Strawberry you train a companion once and everyone on the team gets it.

A worked example from the Next Web: a salesperson trains a companion to research prospects, contact them and enter the details in a sales database. Colleagues then use that companion without starting from scratch.

Other capabilities in the pitch:

• Schedules and triggers. Companions can run on a schedule or start when a new email or sales lead arrives.
• Parallel work. For big jobs, such as researching 200 companies, one companion can split the work across several sub-agents.
• Meeting capture. The browser records Zoom, Microsoft Teams and Google Meet calls locally, so no bot joins the call. It then produces notes and follow-up tasks.

What is actually new in 1.0​

The launch coverage blurs this point. Strawberry's own release history shows that several headline features arrived well before October 26.

Strawberry's 1.0 page lists these as new on launch day:

• Customizable companion appearances.
• Nine ready-made companions, such as Sales Sally and Talent Theo.
• A novelty "Companion Plaza" on the new-tab page.
• A Meetings page that groups recordings, transcripts, summaries, decisions and action items. Transcripts now identify who is speaking.
• A Library that collects what companions produce, including shared team work.

So team sharing is central to the 1.0 pitch, but it did not debut with 1.0. 1.0 is better read as a maturity milestone than as a feature drop.

One small discrepancy in the company's own timeline: it lists open beta on both February 17 and February 22, 2026. The Next Web says downloads opened to everyone in mid-February.

Windows angle, pricing and credits​

Strawberry runs on Windows and macOS. Pricing is credit-based. Per the company's published plans:

• Free: 500 credits a month.
• Intern: $20 a month for 6,000 credits.
• Part-time: $85 a month for 25,000 credits.
• Full-time: $250 a month for 75,000 credits.
• Extra credits: $10 per 1,000.

The company says ordinary browsing uses no credits, while companion chats and agent actions do. Tech.eu confirms the $20, $85 and $250 tiers. These figures are published by the vendor and may change.

The Next Web's "from $20 a month" line is accurate but incomplete. Heavy agent use will push a team into the higher tiers, so budget for credit consumption. Don't budget for the entry price.

Security and oversight: the part IT should read twice​

Handing an agent your logged-in sessions is a big step. Strawberry's stated safeguards, as reported by the Next Web:

• Companions pause for approval before sending emails, publishing posts or changing records.
• Passwords stay in the user's browser.
• Companies can keep customer data within the EU.

Strawberry's changelog also mentions "safety countdowns before risky actions," added in March 2026. That suggests the approval model has been changing during the beta.

These are company representations. The available material doesn't spell out several things admins would want to know:

• The exact data path for each AI action.
• What leaves the device when a companion processes a prompt or a meeting transcript.
• The terms behind the EU-residency claim.
• Whether every consequential action requires approval in every configuration.

Tech.eu says Strawberry is model-agnostic, so the answer likely depends on which AI model provider handles a given task. It also says the agents run on the user's device, which the company ties to European data-sovereignty demands. "Runs on the device" is not the same as "no data leaves the device." Treat the two as separate questions.

Local meeting recording deserves its own policy review. Recording without a visible bot is convenient. It also means other participants may have no visible cue that a call is being captured. Check your organization's consent and retention rules before anyone enables it on Teams calls.

Practical checklist for a pilot​

1. Start with low-risk workflows. Research and drafting are safer than anything that writes to a CRM.
2. Test the approval gates. Confirm they fire for email, posting and record changes in your setup.
3. Review shared companions. A shared companion carries one person's method to everyone, along with any mistakes in it.
4. Check account permissions. A companion acts with the access of the user running it. Least privilege matters more than usual.
5. Get the data terms in writing. That covers meeting content, connected-app data and the EU-residency promise.
6. Track credit burn. Compare actual usage against the plan before committing.

How to read the claims​

Strawberry says it competes less with other AI browsers and more with agent platforms such as Claude, ChatGPT, Grok and Gemini. It is a small company with 12 people. Per the Next Web and Tech.eu, it raised $6 million in October 2025, co-led by General Catalyst and EQT Ventures.

Computerworld reported earlier this year that Strawberry's own tests showed it beating rivals such as Perplexity's Comet and OpenAI's ChatGPT Atlas. It also reported a score of about 78% on the GAIA agent benchmark. Those are vendor-run results. None of the launch coverage offers independent benchmarks or measured productivity gains.

The Next Web ends on the right question. Can an assistant trained by one employee do useful work for a whole team without creating more oversight and maintenance than it saves? Nobody has answered that yet. A free tier makes it cheap to find out, as long as the pilot stays controlled.