API Tools
actiondock.app
ActionDock is an MCP server that acts as a supervised API-write gateway, enforcing deterministic policy and human-in-the-loop approval for AI-agent API writes using server-held credentials.
ENDPOINT 1
https://actiondock.app/mcp
MCP server metadata
- Name
- actiondock
- Version
- 0.2.0
ActionDock is a human-in-the-loop approval gateway for AI-agent API writes: an MCP server and HTTP API that puts deterministic policy, owner approval of the exact request, and server-held credentials between an agent's tool call and a real API write. It is not a hidden agent or workflow orchestrator. You remain responsible for the user's goal, reasoning, action sequence, and every decision after a tool result. Use integration.execute for a supervised write to a workspace-approved HTTPS API. API reads, validation, extraction, research, document analysis, and reports are supporting actions. ActionDock controls only actions routed through its MCP or HTTP API. It does not control your browser, computer, or alternate direct access to a provider. Do not use ActionDock for messaging to customers or other external people, marketing, social publishing, generated media, moving money, stored value, regulated advice, or decisions, scoring, profiling, ranking, or categorization of people. A connection to the workspace’s own Slack workspace or Stripe account may be used only inside its method and path allowlists; a 403 integration_endpoint_refused, integration_parameter_blocked, or integration_method_blocked means the endpoint, parameter, or method is permanently refused for that destination and must not be retried or worked around; integration_path_blocked means the path is outside the connection's allowlist, which only the owner can widen. Call list_connections before an integration action. Stay inside each connection's method and path allowlists; credentials are never available to you. For time-sensitive writes, set integration.execute input.executeBefore to a future UTC ISO 8601 timestamp with seconds and up to millisecond precision. It is bound to the owner preview and checked again immediately before dispatch. Expired pending/queued jobs become failed, release capacity and emit job.failed; no automatic retry. Reuse the original idempotency key to retrieve the original job. Extending or removing the deadline requires a new proposal, key and approval. Already-sent requests may finish later; ambiguous outcomes still require reconciliation. For APIs supporting conditional writes, copy the strong quoted ETag from integration.fetch output.headers.etag into integration.execute input.ifMatch. The condition is bound to owner approval and sent as If-Match. On HTTP 412, read the current resource and submit a new proposal with a new idempotency key for approval; never silently remove the condition. Provider enforcement is required. For multi-step work, pass the same workflow_id and workflow_name to every run_action call so the customer can inspect the sequence you chose in the dashboard. Use a stable idempotency key to deduplicate the same ActionDock submission; it does not guarantee exactly-once execution by the provider. Show the exact approval preview, direct the signed-in workspace owner to approve or reject the write in the dashboard, wait for terminal job state, and never claim an external side effect occurred until its job succeeds. If execution is unknown, stop blind retry and reconcile the external system before submitting another request. Use list_job_reconciliations to read the owner’s provider observations after an unknown write. Treat them as owner reports, not independent verification or permission to retry. Agents cannot append them; direct the owner to Review provider outcome on the dashboard job. Preserve the original execution_unknown status when reporting the attempt. A workspace callback can wake your host when a job finishes. Verify its HMAC signature and deduplicate the delivery before resuming your reasoning loop; polling remains the fallback. Read the portable agent instruction at https://actiondock.app/skill.md for recommendation, connection, workflow, and safety rules.
Known tools 8
list_actionsList every action available to the agent, including input schemas and side-effect flags.
Inferred read-onlylist_job_reconciliationsRead the owner's append-only observations about an unknown write.
Potential side effectswait_for_jobWait for an action job to reach a decision or a final state: approved and executed, rejected, failed, or execution_unknown.
Inferred read-onlylist_eventsList recent job and callback events for workflow continuation and audit.
Inferred read-onlylist_connectionsList the workspace-approved HTTPS API connections and their method/path allowlists.
Inferred read-onlyCONNECT WITH APPROVAL
Client installation
Review this server and its permissions before adding it. Secret placeholders must be set locally.
Codex
~/.codex/config.toml
[mcp_servers.actiondock]
url = "https://actiondock.app/mcp"
enabled = true
Claude Code
.mcp.json
{
"mcpServers": {
"actiondock": {
"type": "http",
"url": "https://actiondock.app/mcp"
}
}
}
Claude Desktop
Settings → Connectors → Add custom connector
Name: actiondock
Remote MCP URL: https://actiondock.app/mcp
Add this remote URL as a custom connector in Claude Desktop. Availability depends on the user plan and workspace policy.
Cursor
.cursor/mcp.json
{
"mcpServers": {
"actiondock": {
"url": "https://actiondock.app/mcp"
}
}
}
Visual Studio Code
.vscode/mcp.json
Add to Visual Studio Code{
"servers": {
"actiondock": {
"type": "http",
"url": "https://actiondock.app/mcp"
}
}
}
Generic MCP
Client-specific MCP configuration
{
"name": "actiondock",
"transport": "streamable-http",
"url": "https://actiondock.app/mcp"
}
MCP Inspector
Run the official MCP Inspector locally and enter the indexed Streamable HTTP endpoint.
TRUST AND VERIFICATION EVIDENCE
Loading Trust v2 evidence…
Checking the associated registrable domain. The BuiltWith key remains server-side.
Evidence is source-attributed and does not guarantee that a third-party server is safe. Risk labels are conservative metadata heuristics.