Strawberry
Also known as: Strawberry Browser, strawberrybrowser.com, Dendrite Systems
Desktop browser for macOS and Windows with built in AI companions that work in the user's signed in tabs. They research, fill forms, source leads and candidates, draft outreach and update connected apps such as Gmail, HubSpot and Ashby. Companions remember context, run routines on schedules or app events, split large jobs across parallel sub agents, and ask before actions the user has not cleared. Nine ready made companions cover sales, recruiting, marketing and more. Free with 500 credits a month, paid plans from $20.
Strawberry is an agentic desktop browser for macOS and Windows. It works as an everyday browser, with tabs, profiles, bookmarks, saved passwords, split view and Chrome extensions, and it imports history and logins from Chrome, Safari, Firefox and Arc. Built into it are AI companions that use the browser itself to do work. A user describes an outcome in plain language, and a companion opens tabs, reads pages, clicks, types and fills forms across the sites the user is already signed into, then hands back a spreadsheet, report, draft or updated record.
Companions are persistent AI teammates, each with its own instructions, files, memory, skills and routines. Users create their own or hire one of nine ready made companions, such as Sales Sally for qualified leads and outreach or Talent Theo for scored candidates, which arrive with the skills and routines for their job. Large tasks are planned in a PLAN.md file and split across sub agents that work in parallel, and every run leaves a work trace of searches, tool calls and browser actions.
Routines start work on a daily, interval or cron schedule or when an app event fires, such as a new email. Companions act in connected apps including Gmail, Slack, Notion, HubSpot, Salesforce, Linear and Ashby, and in any tool reachable through a custom MCP server. Memory carries the user's role, customers, tone and rules into later chats, either shared across companions or kept to one.
Credits pay for companion work, and ordinary browsing costs nothing. The free plan includes 500 credits a month. Intern is $20 a month for 6,000 credits, Part Time $85 for 25,000 and Full Time $250 for 75,000, and top ups at $0.01 a credit roll over. Team adds shared companions, skills and files, owner, admin and member roles, and three free seats, and Enterprise adds pooled credits that roll over and invoicing.
Each user sets per app permissions that decide which actions ask first, run freely or stay disabled, and email approvals can be edited before sending. Strawberry holds a SOC 2 Type 1 report from May 2026 and sells the service for business use. Passwords, history and cookies stay on the device, while chats, synced files and model calls run through its cloud in the EU and the US, with no region choice.
It fits individuals and small teams in sales, recruiting, marketing, consulting and operations who spend their days on research and data entry in web apps and want an assistant that works where they already work, with approval on anything that leaves the building. It fits less well for organizations that need SSO, an audit log across users, a choice of hosting region, or an API for driving agents from their own systems.
Vendor details
Canonical URL
https://strawberrybrowser.com
Category
Browser / computer-use agent
Subcategory
Agentic desktop browser whose AI companions research, fill forms and update apps in the user's signed in session, with routines, memory and per app approvals
Funding status
Independent. Funding is not stated on Strawberry's own pages. It was founded in Stockholm by Charles Maddock, Arian Hanifi and Sebastian Thunman; Dendrite Systems, Inc. in San Francisco operates the service, with Strawberry Browser AB in Stockholm as its EU establishment.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Connected apps through OAuth or API keys, including Gmail, Google Calendar, Drive, Docs and Sheets, Outlook, Slack, Notion, HubSpot, Salesforce, Linear, Ashby, Apollo, QuickBooks, Greenhouse and Klaviyo, among more than 100. Custom tools come in through remote HTTP or local command MCP servers. Any other site is reached through the user's signed in browser, and Chrome extensions, bookmarks and passwords carry over from other browsers.
In practice
Your sales team spends hours pulling prospects from LinkedIn into a sheet. A companion searches Sales Navigator in your signed in browser, checks each account against HubSpot, fills the sheet and drafts outreach in your voice for you to approve before anything is sent.
You want a brief every morning without asking for it. A routine runs on a schedule, reads your inbox and calendar, and leaves a summary with replies drafted for approval, deleting the chat on days when nothing needs you.
A recruiter needs ten candidates for a niche role. Talent Theo researches profiles across the web, scores each one against the role, and checks Ashby so people already in the pipeline are skipped.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
10.0 / 14 capabilities · 71%
| Integrations & Tool Calling | Full |
|---|---|
|
Companions act inside the apps a user connects through OAuth or an API key, including Gmail, Google Calendar, Drive, Outlook, Slack, Notion, HubSpot, Salesforce, Linear, Ashby, Apollo, QuickBooks, Greenhouse and Klaviyo, among more than 100. Each app exposes its own tools, such as reading messages, sending drafted email, updating CRM records or creating tasks. Custom tools come in through remote HTTP MCP servers or local command MCP servers with encrypted environment variables, and sites with no integration are reached through the user's own signed in browser. Sourcedocs.strawberrybrowser.com/integrations/overviewread 2026-10-29 |
|
| Workflow Orchestration | Full |
|
An agent run carries a task through many steps and tabs, searching, reading and comparing pages, using connected app tools, reading and writing workspace files, running code and producing reports, spreadsheets and documents. For bigger jobs the companion first writes a durable PLAN.md with reviewable todos, then spins up sub agents in parallel, each with its own plan in its own session folder, so researching 20 companies can run as 20 agents at once. Interrupted sub agents can be reopened and resumed, and an Agent Control Center shows every running task at a glance. Sourcedocs.strawberrybrowser.com/ai/agent-runsread 2026-10-29 |
|
| Knowledge Grounding & RAG | Partial |
|
Companions work from the files they are given and the pages in front of them. Library holds projects, companion files and generated reports, each companion has its own folder, and a team workspace syncs shared files to every member so companions see team context. Open tabs, connected apps and attached files are read for the task at hand, and research guidance asks for sources and confidence levels on every claim. Files are read when a task needs them, and there is no retrieval index built over a customer's document set. Sourcedocs.strawberrybrowser.com/introduction/what-is-strawberryread 2026-10-29 |
|
| Human Oversight & Guardrails | Full |
|
Each user decides what companions may do without asking. A permission mode sets the default, with Ask Always for actions covered by approval controls, and each connected app lists its actions to set as ask every time, always allow or disabled, either permanently or for one chat. Approval cards show an action before it runs, and email cards let the user change the recipients, subject and body first. A running task can be taken over or stopped, risky actions get a safety countdown, and sub agents ask for decisions when they need them. App reads without their own control run without a prompt. Sourcedocs.strawberrybrowser.com/ai/permissionsread 2026-10-29 |
|
| Security, Identity & Governance | Full |
|
Teams have three roles. Owners control billing, the subscription and deletion, admins manage members, roles, seat plans and shared credits, and members get a read only team view. Per app action permissions and folder access grants govern what each user's companions may do. Strawberry holds a SOC 2 Type 1 report as of 13 May 2026 from Prescient Assurance, shared under NDA, keeps passwords encrypted on the device and encrypts connected app tokens at rest. There is no SSO, SAML or SCIM. Sharing a companion gives the recipient edit access to its files and memories, and app connections are shared across every browser profile on one account. Sourcestrawberrybrowser.com/legal/privacy-policyread 2026-10-29 |
|
| Observability & Auditability | Partial |
|
Every agent run leaves a work trace of its searches, tool calls, file operations, browser actions, sub agent activity, questions and plan progress, which a user can review afterwards to check sources and see how a result was reached. Completed app actions link straight to the document, email, draft or event they created or changed. Routines keep run history per routine and across companions, and each chat shows its own credit usage. Traces live in each user's chats, and team admins get no audit log of what members' companions did. Sourcedocs.strawberrybrowser.com/ai/agent-runsread 2026-10-29 |
|
| Memory & State Persistence | Full |
|
Memory holds the stable context a companion uses in later chats, such as the user's role, company, customers, writing style, target accounts, tool preferences and recurring rules like approval requirements. It comes in two scopes, shared memory every companion reads and memory kept to one companion, and lives as editable files in the companion's synced folder in Library, where outdated facts can be corrected or removed. Companions save new facts as they learn them or when asked. An optional Activity Memory, off by default where offered, logs periodic screenshots of on screen activity and lets the user view, export and delete single entries or all of it. Sourcedocs.strawberrybrowser.com/knowledge/memoryread 2026-10-29 |
|
| Deployment & Data Residency | Not documented |
|
Strawberry is a desktop browser for macOS and Windows. Passwords, history and cookies stay on the device, and the agent works through the user's own signed in browser, while chats, synced files, connected app tokens and model calls run through Strawberry's cloud and its AI providers. Personal data is processed in the European Union and the United States, with EEA data kept in the EEA where feasible, and customers cannot choose a region. There is no private cloud, self hosted or local model option. Sourcestrawberrybrowser.com/legal/privacy-policyread 2026-10-29 |
|
| Prebuilt Agents / Templates / Packs | Full |
|
Nine official companions can be hired ready to work, among them Sales Sally for leads and outreach, Talent Theo for researched and scored candidates, Executive Ellis for morning briefs and follow ups, Developer Dana for testing an app before each release, and Research Rosa for sourced research, alongside Marketing Martin, Consulting Cora, Support Sofia and Operations Olivia. Each arrives with the skills and routines for its job and spends the first chat learning how the user works. A library of Official Skills, filtered by role, covers jobs such as lead sourcing, call preparation, inbox zero, investment memos and bookkeeping, and teams share companions and skills. Sourcestrawberrybrowser.com/updates/strawberry-1-0read 2026-10-29 |
|
| Triggers & Channel Coverage | Full |
|
Routines start work without the user, daily, at an interval, on a cron expression, or when an app event fires, such as a new email in a connected inbox or a new message in a Slack channel. Each routine belongs to a companion and runs as a real chat session, notifying the user when it finds something worth attention and deleting the quiet chats when it does not. Routines can be paused, resumed or run on demand, webhook routines can replay their last run, tasks can be set to run just once, and a companion can turn any successful chat into a routine when asked. Sourcedocs.strawberrybrowser.com/knowledge/routinesread 2026-10-29 |
|
| Model Flexibility & Routing | Full |
|
Paid plans let users choose which AI models a companion uses, add an Expert mode for harder work and give priority access at peak hours, and a user's ChatGPT plan can be connected to save credits. The free plan runs on a curated set of models in Efficient and Default modes. Companion settings also shape which model runs, adaptive thinking is used where a model supports it, and the default model fails over automatically. The full model lineup appears only inside the app. Sourcestrawberrybrowser.com/pricingread 2026-10-29 |
|
| APIs / SDKs / MCP Extensibility | Not documented |
|
Outside systems have no API, SDK or MCP server for starting companions, reading their results or managing routines. MCP support runs the other way, with Strawberry connecting to servers the user adds so companions can call those tools. Chrome extensions install and run in the browser, and routines start from schedules and connected app events, with no public endpoint a customer's own systems can call to hand work to a companion. Sourcedocs.strawberrybrowser.com/integrations/mcp-custom-integrationsread 2026-10-29 |
|
| Testing, Debugging & Optimization | Not documented |
|
Users get no evaluation harness, test cases or quality gate for their companions, skills or routines. A twelve task benchmark suite with an LLM judged scoring rubric compares agentic browsers on research, sourcing and shopping jobs, and Strawberry has run it against its own product and rivals, but it scores browsers in general and cannot check a change to a customer's companion. Research guidance asks for sources and confidence levels and recommends a small sample before a large batch. Sourcestrawberrybrowser.com/benchmarks/specread 2026-10-29 |
|
| Browser / Computer-use | Full |
|
The companion drives the real browser the user works in. It opens tabs, reads pages, clicks, types and fills forms across sites the user is already signed into, including LinkedIn, Sales Navigator and web apps behind a login, and the pages it opens stay grouped inside the chat. A floating controller and a banner show when the companion has the wheel, and the user can watch, take over or stop at any point. Work keeps running in the background while the user browses elsewhere or watches in split view. Sourcestrawberrybrowser.comread 2026-10-29 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Free with 500 credits a month. Intern is $20 a month, Part Time $85 and Full Time $250, with two months free on annual billing.
A subscription per person or per paid seat with a monthly credit allowance, plus prepaid top ups.
Included quota
Free includes 500 credits a month, Intern 6,000, Part Time 25,000 and Full Time 75,000. Each paid Team seat carries its own monthly allowance, and every team gets three free seats, the owner's included. Plan credits refresh monthly and do not roll over, and ordinary browsing never uses them.
What is public
The four individual plans with monthly prices and credit allowances, the top up price, what Team and Enterprise add, and how credits, rollover and rate limits work.
Billing mechanics
Individuals subscribe monthly or annually and spend credits on companion work, with prepaid top ups for more. Teams pay per paid seat by card, each seat with its own allowance, and admins buy shared top ups. Enterprise teams are invoiced against a contract with a pooled credit balance.
Cost watchouts
Every companion reply uses credits, not only agent runs, and long research or large batches can use up a month's allowance quickly. Plan credits do not roll over, and rate limit windows can hold credits back even with allowance left.
Variable cost rationale
The subscription is flat, but credits are spent on every companion reply and more heavily on agent runs, research and app actions, so cost rises with task size and model choice. Spend is capped by the plan allowance and prepaid top ups, and automatic refills run under a monthly spending limit.
Additional watchouts
Plan credits reset monthly without rollover, and team work pauses when no eligible credits remain.
Overage / add-ons
Top ups cost $0.01 a credit ($10 per 1,000), roll over and can refill automatically under a monthly spending limit. Teams have no pay as you go overage, so work pauses when seat, shared and eligible personal credits run out. Short rate limit windows can also hold credits back until they refill.
Sales call required
No, self serve available
Free / trial
Free plan with 500 credits a month in Efficient and Default modes, no card needed
Lowest paid plan
Intern, $20 a month for 6,000 credits
Commercial notes
The service is sold for business use under business terms with a data processing agreement. Paid plans can connect a ChatGPT plan to save credits.
Key ambiguities
Team seat prices and annual plan prices did not render on the pricing page, and credits used per task vary with the job and the model.
Cancellation / refund
Monthly or annual billing, with cancel and reactivate in account settings. Team seat upgrades bill at once with proration and downgrades take effect at period end. Enterprise runs on a contract invoice.
Missing data
Team seat prices, annual plan prices and Enterprise pricing.
Related vendors
- AgentQL — A self healing query language for the web, with natural language…
- AGI, Inc. — AGI, Inc
- Airtop — Cloud browsers for AI agents that compile plain English workflows…
- Asteroid — Healthcare portal integration platform whose supervised browser and…
- Automat AI — AI agents that operate computers visually, sold as Automat Workforce…
- Autotab — General AI agent that learns a workflow from a demonstration and…