Coze Space 扣子空间
ByteDance's free AI agent workspace for research, docs and web tasks
Overview
Coze Space (扣子空间) is ByteDance's agent workspace where 'intern' and domain-expert AI agents research, write reports, build slides and browse the web on your behalf, built on Doubao models and Volcano Engine, distinct from the Coze agent-builder platform.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 12 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
What you can produce with Coze Space 扣子空间
- Autonomous task planning and execution (Exploration + Planning modes)
- Web research and multi-page browsing
- Report, slide deck and document generation
- MCP-based plugin system (60+ integrations: scheduling, email, cloud devices)
- Multi-agent collaboration workspace
- Installable 'Agent Skills' library
ASEAN Perspective
Coze Space 扣子空间 in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Coze Space is ByteDance's answer to Manus and Genspark — an agent you hand a task to and it plans, browses, and produces a deliverable (report, slide deck, spreadsheet) largely unattended. Two operating modes (fast Exploration vs. multi-step Planning) and an MCP-based plugin system covering 60-plus tools give it real range for research and office-adjacent work.
Early reviews comparing it to Manus found the gap real but narrowing: shorter task run-times, fewer pages searched per task, and shallower output depth than Manus at launch. ByteDance has since pushed Coze 3.0 upgrades — multi-agent collaboration, shared project spaces, installable 'Agent Skills' — onto the platform, though it's not always clear which upgrades reach Space specifically versus the separate Coze builder product. It's Chinese-language-first, tightly coupled to Doubao models and Volcano Engine infrastructure, and still carries the rough edges of an invite-gated launch — workable for teams already inside the ByteDance ecosystem, less so as a first choice for ASEAN teams wanting an English-first autonomous agent.
What people say
Coze Space (扣子空间) launched April 18, 2025 as ByteDance's first public test of a general-purpose autonomous agent — hand it a brief and it plans, searches, and produces a finished artifact (report, slide deck, webpage) instead of a chat reply. It runs on Volcano Engine's cloud-native infrastructure with Doubao's 'Deep Thinking' model doing the reasoning, and ships two modes: Exploration for quick single-pass tasks and Planning for multi-step work with checkpoints you can review before it continues.
At launch, hands-on comparisons against Manus — the agent widely credited as the trigger for this category — found real gaps: task runs capped around 5 minutes, roughly 5-10 pages searched per task, and output that read shallower than Manus on the same brief, though Coze Space was free where Manus required an invite code at the time. Access has since flipped: Coze Space opened registration without requiring an invite code, and the invite-code mechanic now works the other direction — completing a task earns you codes to hand out to others.
ByteDance has iterated fast on the broader Coze platform through 2026, rolling multi-agent collaboration, shared project workspaces with persistent memory, and an installable 'Agent Skills' library (one-click modules built by ByteDance and third-party developers) into what it now calls Coze 3.0. It's not always clear from ByteDance's own documentation which of these land in Space specifically versus the separate Coze agent-builder (coze.cn/coze.com) — the two products share branding and underlying infrastructure but serve different jobs, one for building agents, one for delegating tasks to them.
Pricing for Space itself isn't broken out separately in public documentation; the wider Coze ecosystem runs a domestic subscription ladder (free through enterprise, with a July 2026 enterprise-tier price adjustment) and a separate credits system for the international coze.com product. Whether Space draws against the same credit pool or stays free during its extended beta isn't clearly documented, which is worth knowing before you build a workflow around it. The MCP-based extension system — 60-plus plugins spanning scheduling, email, cloud devices and document processors — is the most concrete differentiator versus rival 'AI intern' agents, giving it more built-in reach than most Western competitors offer out of the box.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Coze Space 扣子空间's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Coze Space 扣子空间 unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Coze Space 扣子空间 directly →
Spotted something out of date? Suggest an update →
Coze Space 扣子空间 in the news
Alternatives to Coze Space 扣子空间
More in Agents & Automation
