SharetoBoardTentaClaw tool reference.
A local MCP client starts TentaClaw over stdio. TentaClaw applies local policy, performs the named Poppy operation over outbound HTTPS, and returns a structured result.
Summary
| Tool | Approval class | Poppy effect | Required inputs |
|---|---|---|---|
tentaclaw_version | Local read-only | None | None |
poppy_get_boards | Read-only discovery | Reads policy-visible boards | None |
poppy_get_chats | Read-only discovery | Reads approved chat assistants and nested conversations | board_id |
poppy_ask | Non-read-only | May consume credits; no conversation thread | board_id, chat_id, prompt |
poppy_create_conversation | Non-read-only | Creates a conversation | board_id, chat_id |
poppy_chat | Non-read-only | May consume credits; saves history only when requested | conversation_id, board_id, chat_id, prompt |
tentaclaw_version
Returns the TentaClaw version without contacting Poppy. The local-only candidate also reports transport: stdio and local_only: true. Use it first to confirm the expected local runtime is connected.
poppy_get_boards
Calls Poppy board discovery, then filters the response to the board IDs approved during tentaclaw-mcp init. It is not an account-harvesting tool and should be used only when the approved destination must be resolved.
poppy_get_chats
Requires one exact approved board_id. A board outside local policy returns policy_denied before Poppy is contacted. RC13 preserves the upstream distinction between approved chat assistants and their nested conversations; a conversation name must not be presented as a chat-assistant name.
poppy_ask
Asks one bounded question without creating a persistent conversation thread. Required: approved board, approved chat, and non-empty prompt. Optional fields include model, additional context, plaintext, and a compatibility streaming flag. Requesting streaming fails closed with streaming_not_supported.
When model is omitted, TentaClaw defers to the configured or upstream Poppy default rather than hard-coding a provider model. The Poppy POST request is never retried automatically.
poppy_create_conversation
Creates a Poppy conversation for an approved board/chat pair. Optional name is bounded. This is state-changing and should be used only when continuity is needed. The POST is not automatically retried.
poppy_chat
Continues a known conversation using an encoded single path segment plus approved board/chat IDs. Required: conversation, board, chat, and non-empty prompt. Optional fields include model, context, plaintext, user, source, and save_history. History defaults to false; streaming fails closed; the POST is not retried automatically.
Response and error contract
Tools return a structured success/error envelope. Expected safe errors include configuration_error, policy_denied, invalid_request, request_too_large, response_too_large, redirect_rejected, request_failed, http_error, malformed_response, and streaming_not_supported.
Not exposed by TentaClaw
- Generic HTTP requests, arbitrary Poppy paths, or shell execution
- Arbitrary visual-node construction or connection
- Programmatic chatbot publication
- Browser chatbot URLs as an API execution contract
- MCP over HTTP, OAuth, tunnels, hosted endpoints, or cloud deployment
- Customer identity, billing, metering, throttling, reseller tenancy, or multi-account routing