Integration

Novita AI integration

Novita AI is a cloud provider of AI model APIs and GPU compute. Through this integration an agent can call Novita's language models (OpenAI-compatible chat and text completions, embeddings, reranking), list models with their prices, run batch inference jobs over uploaded JSONL files, generate or layer-split images with the Ming Image models, read asynchronous task results, and manage the account: balance, bills, transactions, rate quotas, team member and API key budgets, and per-key IP and model access policies. GPU cloud management (GPU instances, serverless endpoints, templates, images, networks and prewarm tasks, in both the V1 and V2 GPU APIs) and the passthroughs to other vendors' own protocols (Exa, Tavily, Fish Audio, Bland, Inworld, MiniMax and AutoTrust, most of which have their own integrations) are not exposed. Model calls are billed per token to the Novita account. Authentication is a Novita API key from the key management settings, sent as a Bearer token.

Connect Novita AI How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in Novita AI.Built from the 40 Novita AI tools below, on the connection you add once.

What you can do with Novita AI

Each of these is one operation on one endpoint, available both as a workflow step and as a tool an AI agent can call on its own. Nothing bundles several operations into one action, so what a run did stays readable afterwards.

Each badge shows the default agent permission of an endpoint (allowed: an agent calls it without asking; ask first: you are asked before an agent calls it, when the agent or chat asks permission for sensitive actions; never: no agent may call it), which every user can change endpoint by endpoint in their credential settings.

Most endpoints run on the free plan. A few, mainly the ones that publish on your behalf, need a paid plan.

Batch, 4 operations

cancel_batchCancel batchPOSTAgents: ask firstCancels a batch job that has not finished; requests already processed stay billed. Returns the batch with its new status.
create_batchCreate batchPOSTAgents: ask firstStarts a batch job over an uploaded JSONL file (upload_batch_file) for /v1/chat/completions or /v1/completions; it finishes within 48 hours and every request in it is billed. Follow it with get_batch.
get_batchGet batchGETAgents: ask firstReturns one batch job: status, request counts (total, completed, failed), timestamps, and once done the output_file_id and error_file_id to read with get_file_content.
list_batchesList batchesGETAgents: ask firstLists the account's batch jobs with status, input, output and error file ids, request counts and timestamps.

Key budget, 4 operations

check_key_budgetCheck key budgetPOSTAgents: ask firstTells whether an API key's remaining budget is still sufficient (enough true or false). Read-only.
check_key_budget_for_costCheck key budget for costPOSTAgents: ask firstTells whether an API key's budget is still sufficient after deducting a given cost, in 1/10000 USD (enough true or false). Read-only: nothing is charged.
list_key_budgetsList key budgetsGETAgents: ask firstLists the API key budgets of one team member: key id and name, budget type (Unlimited, One-time, Recurring), limit, used and remaining in 1/10000 USD and the current period. Keys with no budget configured are not listed.
update_key_budgetUpdate key budgetPUTAgents: ask firstSets an API key's spending budget: Unlimited, One-time (a fixed limit) or Recurring (reset Monthly), with budget_limit in 1/10000 USD (at least 10000, $1.00). Changes what the key may spend.

Member budget, 4 operations

check_member_budgetCheck member budgetPOSTAgents: ask firstTells whether a team member's remaining budget is greater than zero (enough true or false). Read-only.
check_member_budget_for_costCheck member budget for costPOSTAgents: ask firstTells whether a team member's remaining budget is at least a given cost, in 1/10000 USD (enough true or false). Read-only: nothing is charged.
list_member_budgetsList member budgetsGETAgents: ask firstLists the team members with their budget: member id, email, phone, role, status, budget type (Unlimited, One-time, Recurring), limit, used and remaining in 1/10000 USD, and the current period. Personal data.
update_member_budgetUpdate member budgetPUTAgents: ask firstSets a team member's spending budget: Unlimited, One-time (a fixed limit) or Recurring (reset Monthly), with budget_limit in 1/10000 USD (at least 10000, $1.00). Changes what the member may spend.

File, 3 operations

delete_fileDelete fileDELETEAgents: ask firstDeletes one file for good (a batch input, output or error file).
get_fileGet fileGETAgents: ask firstReturns one file's metadata: name, size in bytes, purpose, creation and expiry time, and for a batch input the number of requests it holds.
list_filesList filesGETAgents: ask firstLists the uploaded and generated files (batch inputs, outputs and error files): id, file name, size, purpose, creation time and, for output files, expiry.

Key ip policy, 3 operations

delete_key_ip_policyDelete key ip policyDELETEAgents: ask firstRemoves the IP restriction of one API key, so any source IP can call models with it again.
get_key_ip_policyGet key ip policyGETAgents: ask firstReturns the source IPs or IPv4 CIDR ranges allowed to call models with one API key; an empty list means no IP restriction. Management tools are never restricted by it.
set_key_ip_policySet key ip policyPUTAgents: ask firstReplaces the list of source IPs allowed to call models with one API key (at most 100 IPv4 addresses or CIDR ranges); calls from other IPs are refused. Send the full list: read it with get_key_ip_policy first.

Key model policy, 3 operations

get_key_model_policyGet key model policyGETAgents: ask firstReturns which models one API key may call: mode all_enabled (every model the team enabled, minus excludedModels) or selected (only allowedModels), with counts and the last update. A key never configured returns all_enabled.
reset_key_model_policyReset key model policyDELETEAgents: ask firstResets one API key to the default model policy (all_enabled: every model enabled for the team). Only team Owners and Admins may call it.
set_key_model_policySet key model policyPUTAgents: ask firstReplaces which models one API key may call: mode selected with allowedModels, or all_enabled with optional excludedModels. Models must be enabled for the team and available. Only team Owners and Admins may call it.

Model, 2 operations

get_modelGet modelGETAgents: ask firstReturns one language model by id: title, description, context size and the price per million input and output tokens.
list_modelsList modelsGETAgents: ask firstLists the language models Novita serves: id (the value for model), title, description, context size and the price per million input and output tokens. Public, needs no key.

Api key, 1 operation

list_api_keysList api keysGETAgents: ask firstLists the team's API keys: name, numeric id, stringId, masked key (never the plaintext), owner member, expiry, creation and last use time, and with includePolicySummary=true a model access summary per key.

Api key bill, 1 operation

list_api_key_billsList api key billsGETAgents: ask firstLists bills per API key (masked key, key name, product, usage, prices, amounts) from 2026-01-01 on. cycleType Hour allows at most 7 days per call, Day, Week and Month at most 31; longer ranges answer HTTP 400.

Balance, 1 operation

get_balanceGet balanceGETAgents: ask firstReturns the account's balance: available balance, cash top-up balance, credit limit, pending charges and outstanding invoices, each a string in units of 1/10000 USD (10000 is $1.00).

Batch file, 1 operation

upload_batch_fileUpload batch filePOSTAgents: ask firstUploads a JSONL batch input file (purpose batch): one request per line, each with a unique custom_id and a body, all for the same model. Returns the file id for create_batch.

Chat completion, 1 operation

create_chat_completionCreate chat completionPOSTAgents: ask firstGenerates the next assistant message for a conversation with a Novita language model (OpenAI-compatible): returns choices with the message, finish reason and token usage. Billed per input and output token.

Document, 1 operation

rerank_documentsRerank documentsPOSTAgents: ask firstScores a list of documents for relevance to a query with a Novita rerank model: returns each document's index in the input, its relevance score and text, with token usage; top_n keeps the most relevant. Billed per token.

Embedding, 1 operation

create_embeddingsCreate embeddingsPOSTAgents: ask firstTurns texts into embedding vectors with Novita's baai/bge-m3 model: one vector per input, in order, with token usage. Billed per token.

File content, 1 operation

get_file_contentGet file contentGETAgents: ask firstReturns the raw text of a file, typically a batch output or error file in JSONL (one result per line, matched by custom_id).

Fixed term bill, 1 operation

list_fixed_term_billsList fixed term billsGETAgents: ask firstLists subscription (fixed-term) bills, such as monthly GPU or storage purchases and renewals: product, trade type, usage, unit price and amounts. No paging; keep each range within 31 days.

Image, 1 operation

generate_imageGenerate imagePOSTAgents: ask firstGenerates an image from a text prompt with Novita's ming-image-0.1-design model (OpenAI Images protocol). Billed per token; with response_format url the image comes back as a link valid 24 hours.

Image layer, 1 operation

split_image_layersSplit image layersPOSTAgents: ask firstSplits one image into separate layers (such as foreground and background) with Novita's ming-image-0.1-design-layer model, guided by a prompt. Billed per token; returns one PNG per layer.

Monthly bill, 1 operation

list_monthly_billsList monthly billsGETAgents: ask firstLists monthly bills from startMonth on (data starts at 2025-10): month, status (pending, outed, paid, overdue, voided), amounts as USD decimal strings and an invoice download URL. Page with page and pageSize.

Quota, 1 operation

list_quotasList quotasGETAgents: ask firstLists the account's rate quotas (RPM, TPM, IPM) per model or product: current and default limit, per-tier limits and whether an increase can be requested. Filter by modal, quotaType, productType or a model id.

Task result, 1 operation

get_task_resultGet task resultGETAgents: ask firstReturns the state of a Novita v3 asynchronous task (queued, processing, succeed or failed, with progress and the failure reason) and, once it succeeded, its image, video or audio URLs with their lifetime.

Text completion, 1 operation

create_text_completionCreate text completionPOSTAgents: ask firstContinues a text prompt with a Novita language model (OpenAI-compatible legacy completions): returns choices with the text, finish reason and token usage. Billed per input and output token.

Transaction, 1 operation

list_transactionsList transactionsGETAgents: ask firstLists account transactions (top-ups, refunds, consumption): serial number, time, type, channel, amount and wallet balance after it in 1/10000 USD, status and an invoice URL. Page with pageNo and pageSize.

Usage bill, 1 operation

list_usage_billsList usage billsGETAgents: ask firstLists pay-as-you-go bills per Hour, Day, Week or Month for one product category (llm, gpu, serverless, storage, summary): usage, unit prices and USD amounts. No paging; keep each range within 31 days.

Use Novita AI from Claude Code, Cursor or Codex

Connect Novita AI in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its Novita AI operations become tools the client can call, next to every other integration you connected.

Claude Code

claude mcp add --transport http livecontext https://livecontext.ai/mcp --header "X-API-Key: YOUR_API_KEY"

Codex (~/.codex/config.toml)

[mcp_servers.livecontext]
url = "https://livecontext.ai/mcp"

[mcp_servers.livecontext.http_headers]
"X-API-Key" = "YOUR_API_KEY"

Cursor, Claude Desktop and other JSON configs

{
  "mcpServers": {
    "livecontext": {
      "type": "http",
      "url": "https://livecontext.ai/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Set up the MCP server step by step

How to connect Novita AI

Create your workspace
Sign up free. Nothing to install.
Connect Novita AI once
Add your Novita AI connection in Settings. Every tool from this integration reuses it, and secrets are encrypted at rest and visible only to you.
Describe the job in chat
The workflow builds itself in front of you and picks the endpoints it needs. Drag the steps yourself on the canvas whenever you would rather be explicit.