--- title: "Send Message" method: POST path: "/chats/{chatId}/messages" tags: ["Messages"] --- # Send Message `POST /chats/{chatId}/messages` Sends a new message to an existing chat. Blocks until the model response is complete and returns the message response. ## Path parameters - `chatId` string, required — The unique identifier of the chat. ## Request body - object - `message` string, required — The prompt or instruction to send to the model. - `systemPrompt` string — System-level context for the chat, such as frameworks or development environment details. - `modelConfiguration` object — Overrides for the model behavior. - `modelId` 'v0-mini' | 'v0-pro' | 'v0-max' | 'v0-max-fast', required — Model to use for the generation. - `imageGenerations` boolean, required — Enables image generations to generate up to 5 images per version. - `mcpServerIds` string[] — MCP server IDs to enable. When omitted, uses default enabled servers. - `attachments` object[] — Files or assets to include with the message. - `url` string, required — URL of the attachment. - `skills` union[] — Skills to force-attach to the message. Supports skills.sh (`remote`), user/team memory (`memory`), and project (`project`) skills. Maximum 3. - union — A skill to force-attach to the chat. Skills provide domain-specific knowledge to the AI. Use `remote` for skills.sh skills, `memory` for user/team memory skills (including design-system skills), and `project` for skills defined in the chat repo. - object - `type` 'remote', required — Discriminator: a skills.sh skill. - `id` string, required — Skill ID from skills.sh. - object - `type` 'memory', required — Discriminator: a user- or team-scoped memory skill. - `scope` 'user' | 'team', required — Whether the skill lives in user or team memory. - `skillName` string, required — Name of the memory skill to attach. - object - `type` 'project', required — Discriminator: a skill defined in the project repo. - `skillName` string, required — Name of the project skill to attach. ## Response `200` A single message in a chat. - Message — A single message in a chat. - `id` string, required — Unique message identifier. - `chatId` string, required — ID of the chat this message belongs to. - `role` 'user' | 'assistant', required — Who produced this message. - `createdAt` string, date-time, required — ISO timestamp when the message was created. - `updatedAt` string, date-time, required — ISO timestamp when the message was last updated. - `content` string, required — The trailing prose of the message — the agent’s closing summary, or the user’s message text. Empty string when there is no closing prose. - `parts` union[], required — Ordered list of parts that make up the message. Iterate to render the full narrative including thinking, file operations, tool calls, and prose. - union — A single entry in the ordered narrative of a message. Iterate `parts` in order to render the full transcript. - object — A block of text output by the agent. - `type` 'text', required - `text` string, required — Markdown prose written by the agent or user. - `startedAt` string, date-time — ISO timestamp when this part began. - `finishedAt` string, date-time — ISO timestamp when this part completed. - object — Extended thinking output by the agent. - `type` 'thinking', required - `text` string, required — The agent's reasoning trace. - `startedAt` string, date-time — ISO timestamp when this part began. - `finishedAt` string, date-time — ISO timestamp when this part completed. - object — Files read by the agent. - `type` 'file-read', required - `paths` string[], required — The file paths the agent read in this step. - `startedAt` string, date-time — ISO timestamp when this part began. - `finishedAt` string, date-time — ISO timestamp when this part completed. - object — The agent created, modified, deleted, renamed, or patched a file. - `type` 'file-edit', required - `operation` 'create' | 'update' | 'delete' | 'rename' | 'patch', required — The type of file edit performed. - `path` string, required — The target file path. - `toPath` string — The new path when operation is "rename". - `startedAt` string, date-time — ISO timestamp when this part began. - `finishedAt` string, date-time — ISO timestamp when this part completed. - object — The agent searched for something. - `type` 'search', required - `scope` 'repo' | 'web', required — Whether the agent searched the repository or the web. - `query` string, required — The search query. - `startedAt` string, date-time — ISO timestamp when this part began. - `finishedAt` string, date-time — ISO timestamp when this part completed. - object — The agent ran a shell command. - `type` 'bash', required - `command` string, required — The shell command that was executed. - `output` string — Combined stdout/stderr captured from the terminal. May be empty if the command produced no output. - `exitCode` integer, nullable — Process exit code, when available. Reserved for future use; currently always omitted. - `isDangerous` boolean — True if the command was flagged as potentially dangerous and required explicit user approval. - `timeoutMs` integer — Command timeout in milliseconds, when configured. - `startedAt` string, date-time — ISO timestamp when this part began. - `finishedAt` string, date-time — ISO timestamp when this part completed. - object — The agent invoked a tool (MCP, integration, or other dynamic tool). For shell commands, use the "bash" part instead. - `type` 'tool-call', required - `name` string, required — The name of the tool that was invoked (e.g. an MCP tool name or built-in tool identifier). - `input` unknown - `output` unknown - `status` 'ok' | 'error', required — Whether the tool call succeeded or returned an error. - `suggestedPermissions` object[] — Present when the agent is blocked waiting for you to approve this tool call (for example, running a setup or migration script). Pass these objects back unchanged as the `permissions` of a `confirmed-permissions` task to approve. Omitted once the tool has run. - `type` 'ALLOW_DYNAMIC_TOOL_STRICT', required — Permission grant type. - `toolName` string, required — The tool this permission authorizes. - `input` unknown - `taskNameActive` string, nullable — Internal label for the in-progress task. Pass back unchanged. - `taskNameComplete` string, nullable — Internal label for the completed task. Pass back unchanged. - `userMessage` string — Optional message associated with the permission. - `startedAt` string, date-time — ISO timestamp when this part began. - `finishedAt` string, date-time — ISO timestamp when this part completed. - object — A typed escape hatch for agent actions that don’t fit the other part types. Forward-compatible: new action types may appear without API version changes. - `type` 'agent-action', required - `name` string, required — Stable identifier for the action (e.g. "generate_image", "manage_todos", "diagnostics"). See documentation for the registry of known names. - `summary` string, required — A short human-readable summary of what happened. Always populated so clients can render something even without knowing the action name. - `data` union — Structured payload for input-requesting actions. Present on `ask_user_questions`, `exit_plan_mode`, `get_or_request_integration`, and `configure_vercel_connect` parts when the agent is waiting on you; narrow by the part `name`. Omitted for actions that do not carry structured data. - object — Structured payload for an `ask_user_questions` agent action. - `questions` object[], required — The questions the agent is waiting for answers to. Resolve with an `answered-questions` task. - … - object — Structured payload for an `exit_plan_mode` agent action. - `plan` string, required — The full proposed plan, as Markdown. - `summary` string — A concise summary of the plan, when available. - `path` string, required — The path the plan applies to. - object — Structured payload for a `get_or_request_integration` agent action. - `requestedIntegrations` string[], required — Integration names the agent is asking you to connect (for example, "Neon"). Pass these back in `connectedIntegrationNames` when resolving with a `confirmed-steps` task. - `requestedMcpPresets` string[], required — MCP preset names the agent is asking you to add. Pass these back in `connectedMcpPresetNames` when resolving with a `confirmed-steps` task. - union - object — Connector setup is pending. Complete setup at `setupUrl`, poll GET /chats/{chatId}/connect/status until `ready`, then resolve with a `vercel-connect-setup` task. - … - object — The Connect action finished; no user input is needed. - … - `startedAt` string, date-time — ISO timestamp when this part began. - `finishedAt` string, date-time — ISO timestamp when this part completed. - `finishReason` 'stop' | 'length' | 'content-filter' | 'tool-calls' | 'error' | 'other', nullable, required — The reason why message generation finished, including standard completion reasons and error states. - `restorable` boolean, required — True when this is an assistant message that produced restorable code and is not the currently active code. - `attachments` object[] — Files attached to this message. - `url` string, required — URL to the attachment. - `name` string — Original filename, when available. - `contentType` string — MIME type. - `size` integer — Size in bytes. - `authorId` string, nullable, required — ID of the user who authored a user message; null for assistant messages. - `usage` object, required — Token usage and credit cost. All values are zero on user messages and on assistant messages that have not yet generated tokens. - `tokens` object, required — Token counts for this message. - `input` number, required — Prompt input value (non-cached). - `output` number, required — Completion output value. - `cacheRead` number, required — Cache-read input value. - `cacheWrite` number, required — Cache-write input value. - `total` number, required — Sum of input, output, cacheRead, and cacheWrite. - `creditsCost` object, required — Credit cost for this message. - `input` number, required — Prompt input value (non-cached). - `output` number, required — Completion output value. - `cacheRead` number, required — Cache-read input value. - `cacheWrite` number, required — Cache-write input value. - `total` number, required — Sum of input, output, cacheRead, and cacheWrite. ## Other responses - `401` — Response for status 401 - `403` — Response for status 403 - `404` — Response for status 404 - `422` — Response for status 422 - `500` — Response for status 500 --- [API](https://skmtc.net/vercel/apis/v0-platform-api-beta.md) · [All operations](https://skmtc.net/vercel/apis/v0-platform-api-beta/llms.txt) · [OpenAPI document](https://skmtc-service-staging.skmtc.workers.dev/v1/apis/vercel/v0-platform-api-beta/versions/7c7a496f8022/schema)