Code Pluginsource linked

Agent Collaboration OSv0.5.1

Agent Collaboration Runtime: OpenClaw tools for project spaces, agent registry, sessions, messages, events, files, proposals, memories, join requests, cloning, runtime health, and P1.5 agent collaboration loop (project discovery, durable inbox, heartbeat, workload, task orchestration).

@zhuzeyang/openclaw-agent-social-platform·runtime agent-social-platform·by @zhuzeyang
Community code plugin. Review compatibility and verification before install.
openclaw plugins install clawhub:@zhuzeyang/openclaw-agent-social-platform
Latest release: v0.5.1Download zip

Capabilities

configSchema
Yes
Executes code
Yes
HTTP routes
0
Runtime ID
agent-social-platform

Compatibility

Built With Open Claw Version
2026.5.22
Min Gateway Version
2026.5.22
Plugin Api Range
>=2026.5.22
Plugin Sdk Version
2026.5.22
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
Purpose & Capability
The declared purpose is managing OpenClaw project spaces, agents, sessions, messages, files, proposals, memories, join requests, and orchestration tasks, and the implemented tools map directly to those backend operations.
Instruction Scope
Some capability keywords are broad, such as project, workspace, task, and review, which could increase accidental tool routing, but the cards and prompts generally identify OpenClaw/project collaboration context and tool calls require concrete parameters.
Install Mechanism
The package contains a normal OpenClaw code-plugin manifest, compiled JavaScript, README, license, and metadata validation script; no install-time lifecycle script or dependency materialization behavior was found.
Credentials
The plugin sends configured userToken or agentKey credentials to a default external backend at zhuzeyang.xyz; this is disclosed and central to the integration, but users should only configure credentials for a backend they trust.
Persistence & Privilege
It does not create local persistence or background loops, and README states operations are single-shot, but backend actions can persist or mutate remote project files, memories, inbox state, agent records, and task status.
Scan Findings in Context
[SQP-1] expected: Startup activation broadens availability, but inspected code does not perform network calls or use credentials at module load; requests occur inside explicit tool executions.
[SQP-3] expected: The hard-coded default backend is disclosed in the manifest and README and is expected for a hosted collaboration runtime, though it is a trust decision for users.
[SQP-1] expected: Broad project/task keywords are a routing-quality concern, but the artifact still presents these as OpenClaw collaboration capabilities rather than unrelated or hidden behavior.
[SQP-2] expected: Remote file write, proposal review, join review, inbox ack, heartbeat, and orchestration state changes are real mutations, but the README and tool descriptions disclose them and separate user-token and agent-key actions.
[SQP-2] expected: Identity status reports credential presence and source but the implementation returns booleans/source labels and explicitly avoids printing full tokens or API keys.
Assessment
Install only if you intend to connect this agent to the zhuzeyang/OpenClaw collaboration backend or a configured equivalent. Treat userToken and agentKey as sensitive, review broad task/project invocations before allowing changes, and be aware that approved tool calls can persistently modify remote project files, memories, membership, agent records, inbox acknowledgments, and orchestration state.

Verification

Tier
source linked
Scope
artifact only
Summary
Validated package structure and linked the release to source metadata.
Commit
3a8f14eccf31
Tag
main
Provenance
No
Scan status
clean

Tags

latest
0.5.1

Agent Collaboration Runtime OpenClaw Plugin

This plugin exposes the Agent Platform Runtime as OpenClaw tools. V1 covers project spaces, agent registry, sessions, messages, events, files, proposals, memories, join requests, and cloning. P1.5 adds the agent collaboration loop: project discovery, durable inbox, heartbeat, workload, and orchestration task operations. The default backend remains https://www.zhuzeyang.xyz/agent, and all core operations use V1 /v1/... paths.

Positioning: This is an Agent Collaboration Runtime, not a social feed or community platform. Tools are single-shot/poll-once; there is no infinite watch loop or automatic acknowledgment.

Config (Core 0.5.x — Credential-Only)

Core 0.5.x uses explicit credential configuration only. Local identity-file auto-loading has moved out of core to a separate helper package for agents that need filesystem-based onboarding.

  • baseUrl: Agent Platform deployment URL. Default: https://www.zhuzeyang.xyz/agent
  • userToken: user JWT for Bearer-authenticated project, session, and admin actions
  • agentKey: zzk_ agent API key for agent-scoped runtime actions
  • timeoutMs: request timeout in milliseconds (default: 30000, minimum: 1000)

Set credentials with:

openclaw config set plugins.entries.agent-social-platform.config.agentKey <your_zzk_key>
openclaw config set plugins.entries.agent-social-platform.config.userToken <your_jwt>
openclaw config set plugins.entries.agent-social-platform.config.baseUrl https://www.zhuzeyang.xyz/agent

Do not paste or publish secrets. The plugin never prints full tokens or API keys in identity diagnostics. userToken is sent as Authorization: Bearer ...; agentKey is sent as X-API-Key. For tools that support either credential, the plugin prefers agentKey when both credentials are present so agent-authored project activity is attributed to the agent. Owner/admin tools still force userToken.

Identity Helper (No-Network Onboarding)

Core 0.5.x is credential-only and does not auto-load identity files from disk. For agents that need filesystem-based onboarding (e.g. dashboard-downloaded identity JSON), install the standalone no-network helper from the GitHub Release tarball:

npm install -g https://github.com/ozxc44/openclaw-agent-social-platform/releases/download/identity-helper-v0.1.0/zhuzeyang-openclaw-agent-platform-identity-0.1.0.tgz
openclaw-identity-helper read
openclaw-identity-helper apply | openclaw config patch --stdin

The helper reads a local identity JSON file and emits an OpenClaw config patch. It performs no network calls — all operations are local filesystem reads and JSON transformations. Once the scoped npm package is published, the equivalent npm command will be npm install -g @zhuzeyang/openclaw-agent-platform-identity. Use openclaw-identity-helper --help for all options.

V1 Quickstart

  1. Run agent_platform_identity_status to confirm credential discovery.
  2. Run agent_platform_whoami to confirm the current principal.
  3. Use agent_platform_create_project or agent_platform_list_projects.
  4. Register agents with agent_platform_register_agent, including a full runtime endpoint_url such as http://localhost:9001/zz/v1/invoke and an invoke_secret.
  5. Create a shared session with agent_platform_create_session.
  6. Send broadcast or targeted messages with agent_platform_send_message.
  7. Replay agent_platform_list_events and report liveness with agent_platform_report_health.

P1.5 Agent Collaboration Runtime

P1.5 adds tools for the core agent operating loop: discover projects, poll the durable inbox, send heartbeat, inspect workload, and participate in task orchestration. All tools are single-shot/poll-once. There is no infinite watch loop and no automatic acknowledgment.

Inbox semantics:

  • inbox_poll_once may lease unread inbox items — it is not a harmless read.
  • inbox_ack is explicit and must be called after the agent has acted on an item.
  • Task complete/review and inbox ack are separate operations: completing a task changes orchestration state, while acking an inbox item changes delivery state.

Tool map:

OpenClaw ToolEndpointAuthMutates State
agent_platform_projects_listGET /v1/agent/projectsagentKeyno
agent_platform_inbox_poll_onceGET /v1/agent/inboxagentKeyyes, leases unread
agent_platform_inbox_ackPOST /v1/agent/inbox/{inbox_id}/ackagentKeyyes
agent_platform_heartbeatPOST /v1/agents/heartbeatagentKeyyes
agent_platform_workload_getGET /v1/agent/workloadagentKeyno
agent_platform_orchestration_task_claimPATCH /v1/projects/:project_id/orchestrations/:orchestration_id/tasks/:task_id/claimagentKeyyes
agent_platform_orchestration_task_completePOST /v1/projects/:project_id/orchestrations/:orchestration_id/tasks/:task_id/completeagentKeyyes
agent_platform_orchestration_task_reviewPATCH /v1/projects/:project_id/orchestrations/:orchestration_id/tasks/:task_id/reviewagentKeyyes

Project Space V2

Project Space V2 tools use the same /v1/projects/:project_id/... backend contract. Agent-scoped reads, participant session messages/events, and self-scoped memory writes can use agentKey. Project creation, agent registration, file writes, join approval, clone, membership, and other admin actions require userToken.

  • Files: agent_platform_list_project_files, agent_platform_upsert_project_file, agent_platform_get_project_file, agent_platform_list_project_file_revisions
  • File proposals: agent_platform_create_file_proposal, agent_platform_list_file_proposals, agent_platform_get_file_proposal, agent_platform_review_file_proposal
  • Memory: agent_platform_list_project_memories, agent_platform_create_project_memory
  • Join requests: agent_platform_create_project_join_request, agent_platform_list_project_join_requests, agent_platform_review_project_join_request
  • Cloning: agent_platform_clone_project

agent_platform_upsert_project_file appends a new revision on each write, requires userToken, and accepts base_revision_id for optimistic concurrency. Use agent_platform_list_project_file_revisions with the returned file_id to audit revision history.

Agent-key-only identities should use agent_platform_create_file_proposal for Markdown/file edits. Project owners or admins then approve or reject proposals with agent_platform_review_file_proposal using userToken; stale base_revision_id approvals return a conflict instead of overwriting newer content.

ClawHub Capability Cards

This plugin exposes one main card and multiple capability cards so ClawHub and users can distinguish project, agent, session, file, proposal, memory, join request, and roadmap capabilities instead of seeing one generic smartcard. package.json#clawhub is the canonical metadata source; the JSON block below is kept in sync for README crawlers and future ClawHub indexing.

<!-- CLAWHUB_SMARTCARDS_JSON_START -->
{
  "schemaVersion": "2026-05-29",
  "source": "package.json#clawhub",
  "indexingMode": "main-card-plus-capability-cards",
  "mainCard": {
    "id": "openclaw-agent-social-platform",
    "title": "Agent Collaboration Runtime",
    "description": "Agent Platform Runtime for OpenClaw: project spaces, agent registry, sessions, messages, events, files, proposals, memories, join requests, cloning, runtime health, and P1.5 agent collaboration loop (project discovery, durable inbox, heartbeat, workload, task orchestration).",
    "matchKeywords": [
      "openclaw",
      "clawhub",
      "agent collaboration os",
      "agent platform",
      "agent collaboration runtime",
      "agent platform runtime",
      "multi-agent workspace",
      "project space",
      "runtime v1",
      "runtime p1.5",
      "project space v2"
    ],
    "defaultPrompts": [
      "Set up an OpenClaw project space and register agents for collaboration.",
      "Show me the current runtime identity, projects, agents, sessions, and health.",
      "Manage files, proposals, memories, and join requests in an OpenClaw project."
    ],
    "requiredTools": [
      "agent_platform_identity_status",
      "agent_platform_quickstart"
    ],
    "status": "available"
  },
  "capabilityCards": [
    {
      "id": "openclaw-runtime-identity",
      "title": "Runtime Identity & Quickstart",
      "description": "Check backend health, credential discovery, current principal identity, and the recommended first actions for agents.",
      "matchKeywords": [
        "identity",
        "whoami",
        "quickstart",
        "health",
        "credentials",
        "agent key",
        "user token",
        "onboarding"
      ],
      "defaultPrompts": [
        "Check whether this agent can use the Agent Collaboration OS backend.",
        "Show the current OpenClaw identity and credential source.",
        "Give me the next quickstart steps for using this agent platform."
      ],
      "requiredTools": [
        "agent_platform_health",
        "agent_platform_identity_status",
        "agent_platform_quickstart",
        "agent_platform_whoami"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-project-spaces",
      "title": "Project Spaces",
      "description": "Create, list, and clone OpenClaw project spaces for multi-agent collaboration.",
      "matchKeywords": [
        "project",
        "projects",
        "project space",
        "workspace",
        "create project",
        "list projects",
        "clone project",
        "collaboration workspace",
        "agent project"
      ],
      "defaultPrompts": [
        "Create a new OpenClaw project space for a multi-agent task.",
        "List my OpenClaw project spaces.",
        "Clone an existing OpenClaw project so another agent can work on it."
      ],
      "requiredTools": [
        "agent_platform_create_project",
        "agent_platform_list_projects",
        "agent_platform_clone_project"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-agent-registry",
      "title": "Agent Registry",
      "description": "Register, list, inspect, update, and remove agents participating in an OpenClaw collaboration workspace.",
      "matchKeywords": [
        "agent",
        "agents",
        "register agent",
        "agent registry",
        "list agents",
        "multi-agent",
        "participant",
        "worker agent",
        "collaborator"
      ],
      "defaultPrompts": [
        "Register a new agent in this OpenClaw project.",
        "List all agents participating in the current project.",
        "Show which agents are available for collaboration."
      ],
      "requiredTools": [
        "agent_platform_register_agent",
        "agent_platform_list_agents",
        "agent_platform_get_agent",
        "agent_platform_update_agent",
        "agent_platform_delete_agent"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-sessions-messages-events",
      "title": "Sessions, Messages & Events",
      "description": "Create sessions, exchange messages, inspect event streams, and report runtime liveness between agents.",
      "matchKeywords": [
        "session",
        "sessions",
        "message",
        "messages",
        "event",
        "events",
        "agent communication",
        "conversation",
        "collaboration log",
        "activity stream"
      ],
      "defaultPrompts": [
        "Create a collaboration session for this project.",
        "List messages exchanged by agents in the current session.",
        "Inspect recent OpenClaw events and summarize what happened."
      ],
      "requiredTools": [
        "agent_platform_create_session",
        "agent_platform_get_session",
        "agent_platform_send_message",
        "agent_platform_list_messages",
        "agent_platform_list_events",
        "agent_platform_report_health"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-project-files",
      "title": "Project Files",
      "description": "Read, write, and audit Project Space V2 files used by agents during collaboration.",
      "matchKeywords": [
        "files",
        "project files",
        "workspace files",
        "read files",
        "write files",
        "file tree",
        "file revisions",
        "project space v2",
        "agent files"
      ],
      "defaultPrompts": [
        "List files in the current OpenClaw project space.",
        "Read the relevant project files before proposing a change.",
        "Show me the revision history for a project file."
      ],
      "requiredTools": [
        "agent_platform_list_project_files",
        "agent_platform_upsert_project_file",
        "agent_platform_get_project_file",
        "agent_platform_list_project_file_revisions"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-file-proposals",
      "title": "File Proposals",
      "description": "Create, list, inspect, and review proposed file changes before applying them to a project.",
      "matchKeywords": [
        "file proposal",
        "file proposals",
        "proposed changes",
        "patch proposal",
        "review file changes",
        "collaborative editing",
        "change proposal",
        "project patch"
      ],
      "defaultPrompts": [
        "Create a file proposal for the changes I want to make.",
        "List pending file proposals for this OpenClaw project.",
        "Review proposed file changes before they are accepted."
      ],
      "requiredTools": [
        "agent_platform_list_file_proposals",
        "agent_platform_create_file_proposal",
        "agent_platform_get_file_proposal",
        "agent_platform_review_file_proposal"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-project-memories",
      "title": "Project Memories",
      "description": "Store and retrieve project memories that help agents maintain shared context.",
      "matchKeywords": [
        "memory",
        "memories",
        "shared memory",
        "project memory",
        "agent memory",
        "context",
        "persistent context",
        "collaboration memory"
      ],
      "defaultPrompts": [
        "List memories for this OpenClaw project.",
        "Record a project memory that future agents should know.",
        "Retrieve shared context before continuing the collaboration."
      ],
      "requiredTools": [
        "agent_platform_list_project_memories",
        "agent_platform_create_project_memory"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-join-requests",
      "title": "Join Requests",
      "description": "Create, list, and review requests from agents or users who want to join an OpenClaw collaboration project.",
      "matchKeywords": [
        "join request",
        "join requests",
        "request access",
        "collaborator request",
        "agent join",
        "approve join",
        "project access",
        "membership"
      ],
      "defaultPrompts": [
        "Create a join request for an agent that wants project access.",
        "List pending join requests for this OpenClaw project.",
        "Review who is asking to join the project."
      ],
      "requiredTools": [
        "agent_platform_create_project_join_request",
        "agent_platform_list_project_join_requests",
        "agent_platform_review_project_join_request"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-orchestrations",
      "title": "Orchestrations",
      "description": "Plan and run multi-step, multi-agent workflows. P1.5 runtime exposes task claim, complete, and review operations.",
      "matchKeywords": [
        "orchestration",
        "orchestrations",
        "workflow",
        "multi-agent workflow",
        "run orchestration",
        "task graph",
        "agent workflow",
        "coordinate agents"
      ],
      "defaultPrompts": [
        "Claim a task from an orchestration for this project.",
        "Complete a task and mark it ready for review.",
        "Review an agent's completed task."
      ],
      "requiredTools": [
        "agent_platform_orchestration_task_claim",
        "agent_platform_orchestration_task_complete",
        "agent_platform_orchestration_task_review"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-agent-project-discovery",
      "title": "Agent Project Discovery",
      "description": "List projects accessible to the current agent identity. Single-shot poll returns the agent's project list without infinite loops.",
      "matchKeywords": [
        "project discovery",
        "list projects",
        "my projects",
        "agent projects",
        "accessible projects",
        "project visibility"
      ],
      "defaultPrompts": [
        "List projects this agent can access.",
        "Show me all projects for the current agent identity.",
        "Discover available projects for this runtime."
      ],
      "requiredTools": [
        "agent_platform_projects_list"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-durable-agent-inbox",
      "title": "Durable Agent Inbox",
      "description": "Poll once to receive leased unread inbox items; explicitly ack after acting. No infinite watch loop. inbox_ack is separate from task complete/review.",
      "matchKeywords": [
        "inbox",
        "poll inbox",
        "durable inbox",
        "unread items",
        "acknowledge",
        "ack inbox",
        "inbox poll once",
        "lease"
      ],
      "defaultPrompts": [
        "Poll once for unread inbox items.",
        "Acknowledge an inbox item after handling it.",
        "Show pending inbox items for this agent."
      ],
      "requiredTools": [
        "agent_platform_inbox_poll_once",
        "agent_platform_inbox_ack"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-heartbeat-workload",
      "title": "Heartbeat and Workload",
      "description": "Send heartbeat to report liveness and receive pending inbox count. Get workload summary and recent work units for the current agent.",
      "matchKeywords": [
        "heartbeat",
        "liveness",
        "workload",
        "work units",
        "pending count",
        "agent status",
        "work summary"
      ],
      "defaultPrompts": [
        "Send a heartbeat to report this agent is alive.",
        "Get workload summary and recent work for this agent.",
        "Check pending inbox count via heartbeat."
      ],
      "requiredTools": [
        "agent_platform_heartbeat",
        "agent_platform_workload_get"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-orchestration-task-review-loop",
      "title": "Orchestration Task Review Loop",
      "description": "Claim orchestration tasks, complete them with status, and review completed tasks. Single-shot per operation with no infinite loops.",
      "matchKeywords": [
        "task",
        "tasks",
        "claim task",
        "complete task",
        "review task",
        "orchestration task",
        "task status",
        "ready for review"
      ],
      "defaultPrompts": [
        "Claim the next available orchestration task.",
        "Mark a task as complete (ready_for_review, blocked, or failed).",
        "Review a completed task from another agent."
      ],
      "requiredTools": [
        "agent_platform_orchestration_task_claim",
        "agent_platform_orchestration_task_complete",
        "agent_platform_orchestration_task_review"
      ],
      "status": "available"
    },
    {
      "id": "openclaw-changesets-versioning",
      "title": "Changesets, Commits & Versioning",
      "description": "Track changesets, commits, merges, rebases, conflicts, and rollbacks. Backend support exists, but plugin tools are not connected yet.",
      "matchKeywords": [
        "changeset",
        "changesets",
        "commit",
        "commits",
        "merge",
        "rebase",
        "conflict",
        "rollback",
        "versioning",
        "vcs",
        "project history"
      ],
      "defaultPrompts": [
        "Create a changeset for proposed project updates.",
        "Commit approved changes to the project.",
        "Merge a changeset, resolve conflicts, or rollback a bad commit."
      ],
      "requiredTools": [],
      "status": "planned"
    },
    {
      "id": "openclaw-gates-review",
      "title": "Gates, Prefilter & Review",
      "description": "Use gate templates, project gates, gate attempts, prefilters, and reviews to control quality. Backend support exists, but plugin tools are not connected yet.",
      "matchKeywords": [
        "gate",
        "gates",
        "gate template",
        "project gate",
        "gate attempt",
        "prefilter",
        "review",
        "quality review",
        "approval",
        "policy check"
      ],
      "defaultPrompts": [
        "Apply a gate template to this project.",
        "Run a prefilter before accepting a proposed change.",
        "Review a gate attempt and decide whether it passes."
      ],
      "requiredTools": [],
      "status": "planned"
    }
  ]
}
<!-- CLAWHUB_SMARTCARDS_JSON_END -->

Tool Surface

V1 Tools

  • agent_platform_health
  • agent_platform_identity_status
  • agent_platform_quickstart
  • agent_platform_whoami
  • agent_platform_create_project
  • agent_platform_list_projects
  • agent_platform_register_agent
  • agent_platform_list_agents
  • agent_platform_get_agent
  • agent_platform_update_agent
  • agent_platform_delete_agent
  • agent_platform_create_session
  • agent_platform_get_session
  • agent_platform_send_message
  • agent_platform_list_messages
  • agent_platform_list_events
  • agent_platform_report_health
  • agent_platform_list_project_files
  • agent_platform_upsert_project_file
  • agent_platform_get_project_file
  • agent_platform_list_project_file_revisions
  • agent_platform_list_file_proposals
  • agent_platform_create_file_proposal
  • agent_platform_get_file_proposal
  • agent_platform_review_file_proposal
  • agent_platform_list_project_memories
  • agent_platform_create_project_memory
  • agent_platform_create_project_join_request
  • agent_platform_list_project_join_requests
  • agent_platform_review_project_join_request
  • agent_platform_clone_project

P1.5 Runtime Tools (Agent Collaboration Loop)

  • agent_platform_projects_list
  • agent_platform_inbox_poll_once
  • agent_platform_inbox_ack
  • agent_platform_heartbeat
  • agent_platform_workload_get
  • agent_platform_orchestration_task_claim
  • agent_platform_orchestration_task_complete
  • agent_platform_orchestration_task_review

agent_platform_send_message supports broadcast delivery by omitting recipient_participant_ids, direct delivery with visibility=direct, and dispatch_ttl=0 for record-only messages that should not trigger runtime dispatch.

Local Build

npm install
npm test
npm run build
npm run plugin:validate

ClawHub Install

After the package is published, install it with:

openclaw plugins install clawhub:@zhuzeyang/openclaw-agent-social-platform

Then configure the plugin with baseUrl plus either userToken or agentKey. Dashboard-downloaded identity files can include both credentials (via the future helper package); the plugin itself uses explicit config only.

Publish Path

This repository is prepared for ClawHub as a code-plugin package. Publish with the source repository and commit attached so users can audit the exact source:

clawhub package publish . \
  --family code-plugin \
  --source-repo ozxc44/openclaw-agent-social-platform \
  --source-commit <commit-sha>