How to Set Up Shared Workspaces in OpenClaw

OpenClaw workspace setup configures shared storage where multiple OpenClaw agents access files persistently across sessions. Agents generate reports, datasets, images, or code that require organized storage, version history, previews, and collaboration tools.

Fast.io integrates via the ClawHub skill, delivering 14 dedicated tools for file management, sharing, and AI queries. Unlike local storage that vanishes on restart or email limited to 25 MB attachments per Gmail's policy, Fast.io offers browser previews for videos, PDFs, and images, plus real-time comments and activity feeds.

Consider a data pipeline: Agent A processes raw data into CSV files. Agent B runs analysis and stores charts. Agent C compiles a client-ready package. All files live in one searchable workspace. Semantic search locates "Q3 sales analysis charts" by meaning, not filename. Agents can easily transfer ownership to humans.

This setup gives agents persistent memory to run complex workflows without data loss. Workspaces support folders, locking for concurrent access, and webhooks for event-driven automation.

Verify these before starting openclaw workspace setup:

1. OpenClaw installed and operational on your machine. Clone from GitHub if needed: git clone https://github.com/steipete/openclaw.
2. Terminal access with npx (bundled with npm) or ClawHub CLI.
3. Modern browser for Fast.io authentication.
4. Sample OpenClaw agent script testing tool calls.
5. No prior Fast.io account required; authentication creates one on the free agent plan with 50 GB storage and 5,000 credits per month.

Quick ClawHub test: clawhub list. Base tools appear; Fast.io tools added post-install. Reference ClawHub docs for CLI details and Fast.io MCP guide for tool specs.

Optional: Review OpenClaw examples for file-handling prompts.

Execute in terminal:

npx clawhub@latest install fast-io

Alternative:

clawhub install dbalve/fast-io

Installation registers 14 tools: auth, workspace management, storage operations, sharing, AI chat, events, members, and more. Zero configuration. No env vars or files.

Initial agent tool call triggers browser for signup/authentication. Use any email; agent plan activates automatically (50 GB, 5,000 credits/month, 1 GB max upload).

Verify: Prompt agent "List Fast.io tools." Output includes upload, list_files, create_share, ai_query, etc.

Give Your AI Agents Persistent Storage

Free 50 GB, 14 tools, semantic search. Agents collaborate with humans right away.

Get Started Free

Agents auto-create orgs/workspaces on first tool use, or prompt explicitly:

1. Authenticate if new.
2. Call workspace tool: name "TeamProject".
3. Toggle Intelligence Mode for semantic search.

Sample prompt: "Create workspace 'MarketingCampaign', upload sample.csv, enable AI indexing."

Invite collaborators: Generate links/IDs for agents/humans. Workspaces handle folders, versioning, previews (PDF/video/images up to 1 GB).

Code snippet (agent pseudocode):

tools = clawhub.tools()
ws = tools.workspace.create(name="SharedAgents", intelligence=True)
invite = tools.member.invite(ws.id, role="member")
print(f"Invite other agents: {invite.link}")

Each agent installs skill independently, authenticates separately, joins via invite ID.

Expanded example:

### Agent A creates
ws = tools.workspace.create("DataPipeline")

### Agent B joins
ws = tools.workspace.get("DataPipeline-ID")
tools.member.add(ws.id, invite_id="...")

### Concurrent work with locks
lock = tools.storage.lock(ws.id, file_id="data.csv")
### process...
tools.storage.unlock(ws.id, file_id="data.csv")

Intelligence Mode auto-indexes for queries. Webhooks alert on changes: tools.webhook.subscribe(ws.id, "file_uploaded").

Handoff: tools.org.transfer(ws.org_id) generates claim link for humans.

Define clear tool contracts and fallback behavior so agents fail safely when dependencies are unavailable. This improves reliability in production workflows.

ClawHub skill bundles these into natural language calls:

1. auth: signin/signup, status, api-key manage.
2. org: create/list, transfer ownership.
3. workspace: create/list/update, intelligence toggle.
4. member: invite/add/remove.
5. storage: list/move/copy/delete/search/lock.
6. upload: text/binary/web-import.
7. share: create Send/Receive/Exchange.
8. comment: add/list on files.
9. ai: chat_with_files, scope queries.
10. event: activity-poll, changes.
11. user: invitations accept.
12. download: workspace/share files.
13. preview: generate URLs/transforms.
14. metadata: templates/extract/query.

Full docs: MCP skill.md.

Practices:

• Use one workspace per project.
• Turn on Intelligence Mode for search-heavy workflows.
• Use locks for multi-agent edits.
• Use webhooks instead of polling.
• Transfer ownership if credits run out (5,000/month).

Troubleshooting:

• Auth loop: Clear cache, retry signup.
• 402 error (credits exhausted): List usage auth status, transfer org.
• Large files fail: Under 1 GB limit.
• Invite errors: Verify ID format.
• Indexing delay: Poll ai_state in file details.

Scale: URL import from Drive/Dropbox, ownership transfer.

Capture these lessons in a shared runbook so new contributors can follow the same process. Consistency reduces regression risk and makes troubleshooting faster.

Example 1: Data Pipeline Agent1: Upload raw data, lock dataset.csv. Agent2: Analyze, append results. Agent3: Generate report PDF, create Send share.

Prompt chain: "Pipeline: process sales.csv → charts.png → summary.pdf → share with client@email.com."

Example 2: Human-AI Review Agent builds prototype files. Invite human, add comments. Agent reads comments via comment list, iterates.

Example 3: Knowledge Base Build Import dozens of PDFs via URL. Enable intelligence. Query: "Summarize compliance risks across docs."

Define clear tool contracts and fallback behavior so agents fail safely when dependencies are unavailable. This improves reliability in production workflows.

Fast.io works well for agent teams: native MCP, RAG, handoff support.