Validated against primary sources, Feb 25, 2026. Reframed: competition assessed through the lens of “agent-human collaboration,” not just “agent storage.”
| Attribute | Detail |
|---|
| What it does | ”Workspaces for Agentic Teams” — shared workspace where agents collaborate |
| Founders | MediaFire founders (deep file-sharing experience) |
| Product | LIVE — agent accounts, shared workspaces, 251 MCP tools, semantic search, built-in RAG |
| Free tier | 50GB free |
| Strengths | Right vision, relevant founders, live product, generous free tier |
| Gap | Focuses on agents-as-users. Human collaboration (viewing, commenting, real-time dashboards) is not the primary use case. |
| Threat level | HIGH |
Our differentiation: Fast.io builds workspaces for agents. We build workspaces for agents AND humans to collaborate. The human dashboard, activity feed, comments, and feedback loop are our wedge.
| Attribute | Detail |
|---|
| What | SQLite-based filesystem for AI agents |
| GitHub | 134 stars, v0.4.1 |
| Gap | No agent identity, no human dashboard, no collaboration, no enterprise features |
| Threat level | MEDIUM — good tech building block, not a collaboration product |
| Attribute | Detail |
|---|
| What | Sandboxed cloud environments (Firecracker microVMs) |
| Funding | $14.4M |
| Gap | Compute, not collaboration. No shared workspaces, no human visibility |
| Threat level | LOW-MEDIUM — complementary (compute + our workspace) |
| Attribute | Detail |
|---|
| What | ”Give every agent a computer” — sandboxes with state persistence |
| Funding | $24M Series A, $2M ARR |
| Gap | Compute-centric. Workspace hours expire. No human collaboration |
| Threat level | LOW-MEDIUM — complementary |
Duvo is a retail operations automation company (Index Ventures $15M). NOT agent file storage. See prior correction.
| Attribute | Detail |
|---|
| Status | Generally available |
| Integrations | Claude, Copilot Studio, Azure API Center, Mistral |
| Revenue | Box: $1.1B/year |
| Gap | Human storage retrofitted for agents. No agent identity. Optimized for agents READING enterprise content, not agents + humans COLLABORATING on new work. |
| Threat level | HIGH for enterprise — existing contracts, trusted brand |
| Attribute | Detail |
|---|
| What | No-code agent builder + cross-platform content search |
| Gap | User-created agents, not agent-native. Agents are extensions of human users, not first-class collaborators. No agent identity layer. Optimized for human-to-human collaboration, not agent-human collaboration. |
| Threat level | HIGH — “good enough” for many, billion-user platform |
| Attribute | Detail |
|---|
| What | Select files → create .agent file → Copilot experience |
| Limits | Max 20 files per agent |
| Gap | Limited scope. No agent-to-agent sharing. No autonomous agent identity. One-directional (human → agent), not collaborative. |
| Threat level | HIGH for enterprise — M365 distribution |
| Attribute | Detail |
|---|
| Gap | Basic CRUD only. No agent identity, no collaboration, no human dashboard, no activity feed. “Good enough” for single-agent prototypes. |
| Threat level | MEDIUM — breaks at multi-agent and human-oversight scale |
| Component | What | Relevance to Us |
|---|
| OpenClaw (215K+ stars) | Self-hosted autonomous AI agent | Primary beachhead user — always-on agents need persistent shared workspace |
| ClawHub (5,705+ skills, 1.5M downloads) | Skill registry/marketplace | Primary distribution channel — publish Skill for instant access to 100K+ instances |
| MoltBook (2.66M agents, 17K owners) | Social network for AI agents | Demand validation — proves agents need platforms, humans want to observe |
| Cline (5M+ installs, 57K stars) | VS Code coding agent | Secondary distribution — Cline Marketplace listing for one-click install |
| Roo Code / Kilo Code | Cline forks | Additional distribution — MCP-compatible, same config format |
The OpenClaw opportunity: 100K+ autonomous agents running 24/7 with zero human visibility. 7 open GitHub issues describe exactly what we build. No native workspace solution exists. Fast.io is the only third-party option, but it lacks human dashboards.
| Company | What | Relationship to Us |
|---|
| Browserbase ($67.5M) | Browser infra for agents | Complementary — agents browse, then save to workspace |
| AgentMail (YC S25) | Email for agents | Parallel primitive — we compete via Mailmolt |
| Keycard ($38M) | Agent identity | Complementary — identity layer, no services |
| CrowdStrike/SGNL ($740M) | Agent identity/auth | Complementary — enterprise security |
| Notion / Coda | Human collaboration docs | Adjacent — human-first, no agent integration |
| Mem0 | Persistent memory for OpenClaw | Adjacent — memory layer, not file workspace. No human visibility. |
| MoltBook | Agent social network | Complementary — proves demand, not competitive (social ≠ workspace) |
| Capability | Our Vision | Fast.io | Box MCP | Google WS | MS OneDrive | MCP Wrappers |
|---|
| Agent-native design | Yes | Yes | No | No | No | No |
| Human collaboration | Yes | Partial | No | Human-only | Human-only | No |
| Human dashboard | Yes | Unknown | No | Human-only | Human-only | No |
| Activity feed | Yes | Unknown | No | No | No | No |
| Comments/feedback | Yes | No | No | Human-only | Human-only | No |
| Persistent storage | Yes | Yes | Yes | Yes | Yes | Yes |
| Agent identity | Yes | Yes | No | No | No | No |
| Cross-agent sharing | Yes | Yes | Limited | Limited | No | No |
| Enterprise audit | Yes | Unknown | Yes | Yes | Yes | No |
| Suite integration | Yes | No | Yes (Box) | Yes (Google) | Yes (M365) | No |
| Live product | No | Yes | Yes | Yes | Yes | Yes |
- Agent-human collaboration on shared files — agents and humans as co-workers, not agents-as-tools
- Human dashboard with activity feed — see what agents are doing in real-time
- Comment/feedback loop — humans annotate, agents read feedback
- Cross-platform — any MCP-compatible agent (not locked to Google/Microsoft)
- Suite integration — mail + workspace + skills, unified identity
- Fast.io could add human collaboration features (it’s engineering, not innovation)
- Google/Microsoft could build agent-native collaboration (they own the collaboration market)
- The concept of “agent-human collaboration” may be ahead of market demand
- We have no live product; everyone else does
- 12-18 months before incumbents optimize for agent-human collaboration
- Google/Microsoft are adding agents TO human tools (agent-as-extension)
- We’re building collaboration FROM SCRATCH for agent-human teams (collaboration-as-product)
- These are fundamentally different approaches — ours is better for agent-heavy workflows, theirs is better for human-heavy workflows with occasional agent help