Claude visitors: new accounts receive $1.00 in bonus API credit within 30 days. One per person; usage credits only.
Use RunAPI in WorkBuddy.
WorkBuddy is Tencent's AI agent workbench and speaks the MCP protocol natively. Add RunAPI as an MCP server, drop your key in RUNAPI_API_KEY, point the server at https://runapi.ai/mcp, and WorkBuddy can call 250+ models without leaving the workbench.
Configure this WorkBuddy install to use RunAPI over MCP.
Requirements:
- Read the API key from RUNAPI_API_KEY. Do not hardcode the key.
- Add an MCP server named runapi.
- Use url "https://runapi.ai/mcp".
- Use the streamable HTTP transport.
- Send the header "Authorization: Bearer $RUNAPI_API_KEY".
- After adding the server, list the tools it exposes and report them.
- Run a short test generation to confirm the connection works.
- If RUNAPI_API_KEY is missing, ask me to set it before adding the server.{
"mcpServers": {
"runapi": {
"url": "https://runapi.ai/mcp",
"headers": {
"Authorization": "Bearer ${RUNAPI_API_KEY}"
}
}
}
}export RUNAPI_API_KEY="runapi_xxx"
# Open WorkBuddy Settings > MCP Servers
# Verify "runapi" server shows connectedWhat WorkBuddy is, and how RunAPI extends it
WorkBuddy is Tencent's AI agent desktop workbench, part of the CodeBuddy suite. It ships with 14+ native models and a SkillHub of 70,000+ skills, and it speaks the MCP protocol so you can plug in external tool servers. Out of the box it covers chat, coding, and office tasks, but its model choice is limited to what Tencent bundles.
RunAPI removes that ceiling. Connect it as a single MCP server and WorkBuddy can reach 250+ AI models from 26 providers across 5 modalities — text, image, video, music, and speech — through one API key and one endpoint. No per-provider signups, no juggling keys — add the server once and every model shows up as a callable tool inside the workbench.
Why connect RunAPI to WorkBuddy
MCP one-click connect
WorkBuddy speaks MCP natively, so RunAPI registers as a single server. Paste the prompt once and every model becomes a callable tool — no plugins or bridges.
250+ models, one key
Reach 250+ models from 26 providers with a single RUNAPI_API_KEY. Switch between GPT, Claude, Gemini, Flux, Kling, and Suno without adding new accounts.
5 modalities
Generate text, images, video, music, and speech from the same MCP server. WorkBuddy handles the task; RunAPI routes it to the right model automatically.
OpenAI compatible
Chat and completion models follow the OpenAI-compatible request shape, so WorkBuddy's existing tool calls and prompts work without rewrites.
Pay-per-use pricing
No seat licences or monthly minimums. You pay only for what each generation costs, and pooled billing covers every model behind the single key.
Enterprise security
Keys stay in the environment, never hardcoded. Requests are scoped to your account, and usage is auditable per key from the RunAPI dashboard.
Add RunAPI to WorkBuddy with a prompt
Paste the setup prompt
Copy the prompt from the first tab into WorkBuddy. It tells WorkBuddy to register RunAPI as an MCP server and read the key from RUNAPI_API_KEY.
Copy prompt
Review the MCP server config
WorkBuddy should add an MCP server named runapi pointing at https://runapi.ai/mcp over streamable HTTP, with the Authorization header carrying your RunAPI key.
url: https://runapi.ai/mcp
Verify the connection
Ask WorkBuddy to list the tools exposed by the runapi MCP server, then run a short test generation before using it for a larger task.
List MCP tools
What you can build from the workbench
Office documents
Draft reports, summaries, and slides inside WorkBuddy, then let RunAPI route the heavy reasoning to a top-tier chat model for cleaner long-form output.
GPTData visualization
Ask WorkBuddy to analyze a dataset and generate charts or narrative insights, with Claude handling structured reasoning over large tables.
ClaudeVideo generation
Turn a storyboard or product brief into short clips without leaving the workbench — RunAPI runs the async video job and returns the finished URL.
KlingMusic and audio
Generate background tracks or jingles from a text brief. WorkBuddy sends the request; RunAPI polls Suno and hands back the finished audio.
SunoDesign assets
Create logos, hero images, and marketing visuals on demand. Flux renders from a prompt and the image URL lands straight back in WorkBuddy.
FluxTranslation and localization
Localize docs, UI strings, and support replies at scale using a fast, low-cost model so batch translation stays affordable.
DeepSeekWhy teams add RunAPI to WorkBuddy
One integration, every model
Instead of wiring up each provider's SDK, you register one MCP server. New models added to RunAPI appear in WorkBuddy without any extra setup on your side.
No key sprawl
A single RUNAPI_API_KEY replaces the pile of OpenAI, Anthropic, and Google keys you would otherwise manage, rotate, and secure separately.
Predictable pooled billing
All model usage rolls up to one balance and one invoice, so finance sees a single line item instead of a dozen provider bills.
Future-proof by default
RunAPI tracks new releases across providers. When a stronger model ships, it is available through the same endpoint — no migration work in WorkBuddy.
WorkBuddy native vs. WorkBuddy with RunAPI
| Feature | WorkBuddy Native | WorkBuddy + RunAPI |
|---|---|---|
| Model count | 14+ bundled models | 250+ models across 26 providers |
| Modalities | Mostly text and code | Text, image, video, music, and speech |
| Adding a provider | Wait for native support | Already included — no new setup |
| API keys to manage | One per external provider | A single RUNAPI_API_KEY |
| Billing | Separate per provider | One pooled, pay-per-use balance |
MCP server fields
| Field | Value | Purpose |
|---|---|---|
mcp_servers.name |
runapi |
Registers a named MCP server that WorkBuddy can call for model tools. |
url |
https://runapi.ai/mcp |
Points WorkBuddy at RunAPI's MCP endpoint for model discovery and tasks. |
transport |
streamable_http |
Uses the HTTP streaming transport WorkBuddy expects for remote MCP servers. |
auth_header |
Authorization: Bearer $RUNAPI_API_KEY |
Sends the RunAPI key from the environment instead of hardcoding it. |
One MCP server for images, video, and music.
WorkBuddy ships strong chat and coding models, but image, video, and music usually mean extra tools or external services. RunAPI replaces all of that — the same runapi MCP server and RUNAPI_API_KEY handle image, video, and music requests alongside chat, with no plugins or GPU rentals.
Image Generation
Call Flux Kontext, Imagen 4, Seedream, GPT Image, and Nano Banana through the RunAPI MCP server. No local GPU, no workflow graphs — send a prompt and receive an image URL.
Browse image modelsVideo Generation
Generate video from text or images with Kling 3.0, Seedance 2.0, Veo 3, and Wan 2.7. RunAPI handles async polling and callbacks with the same task format across all video models.
Browse video modelsMusic Generation
Access Suno v4 through v5.5 directly through RunAPI. No third-party bridge service needed — pass lyrics, style, and duration, and poll for the finished audio.
Browse music modelsUse any model in WorkBuddy
Each guide shows the prompt, curl command, parameters, and FAQ for calling a specific model through RunAPI inside WorkBuddy.
InfiniteTalk
Talking avatar video from audio
Grok Imagine
xAI image and video generation
Flux Kontext
Text-guided image editing
Z Image
Fast bilingual image generation
Topaz
AI image and video upscaling
Imagen 4
Google image generation
Seedream
Bytedance 4K image generation
Qwen 2 Image
Alibaba image generation and editing
Kling
Video generation with motion control and sound
Veo 3
Google cinematic video with native audio
Seedance
Budget video generation by Bytedance
Hailuo
MiniMax video generation
Wan
Open-source video and image generation
Runway
Creative video with camera control
Luma
Spatial-aware video generation
Suno
AI music generation from text
ElevenLabs
Text-to-speech and voice AI
GPT
OpenAI chat and responses API
Claude
Anthropic messages API
DeepSeek
Budget LLM with reasoning
Gemini
Google AI models, no GCP needed
Nano Banana
Google image generation with text rendering
Midjourney
Midjourney image generation, editing, and image-to-video
Ideogram V3
Typography-focused image generation
Recraft
Vector art and image upscaling
Flux 2
BFL image generation, pro and flex
GPT Image
OpenAI image generation and editing
WorkBuddy and RunAPI questions
WorkBuddy speaks MCP, so you add RunAPI as an MCP server. Paste the setup prompt, which registers a server named runapi pointing at https://runapi.ai/mcp and reads your key from RUNAPI_API_KEY. Verify by asking WorkBuddy to list the tools the server exposes.
WorkBuddy bundles 14+ models, mostly for chat and code. RunAPI adds 250+ models across 26 providers and 5 modalities through one key, so you can generate images, video, music, and speech from the same workbench without waiting for native support.
SkillHub is WorkBuddy's library of 70,000+ agent skills — workflows and automations that decide what to do. RunAPI is the model layer underneath that decides which AI does the work. You keep using SkillHub for orchestration and let RunAPI supply the 250+ models any skill needs to run.
Creating an account and API key is free. API calls use prepaid, pay-as-you-go billing; add funds before making requests.
Yes. Your key lives in the RUNAPI_API_KEY environment variable, never hardcoded in the config. Requests are scoped to your account, every call is auditable per key from the dashboard, and you can rotate or revoke the key at any time without touching WorkBuddy's other settings.
Yes. Chat and completion models use the OpenAI-compatible request shape, and the MCP server exposes each model as a standard tool that WorkBuddy already knows how to call. Your existing WorkBuddy prompts, skills, and automations keep working without any rewrites or migration effort.
Yes. The same runapi MCP server exposes video models like Kling and Veo and music models like Suno alongside chat and image tools. WorkBuddy sends the request, RunAPI handles the async polling and callbacks, and the finished URL comes back into the workbench with no extra services to wire up.
Set RUNAPI_API_KEY in the environment that launches WorkBuddy — your shell profile, service unit, or the app's environment settings. Desktop launches do not always inherit interactive shell variables, so confirm the key is visible to the process before the MCP server tries to authenticate a request.
No. That is the point of RunAPI. One RUNAPI_API_KEY covers every model behind the MCP server, so you never manage separate OpenAI, Anthropic, or Google keys. Usage from all 26 providers pools into a single balance with one invoice, so there is nothing extra to rotate or reconcile.
RunAPI adds new models across providers as they ship, and they appear through the same MCP endpoint. WorkBuddy sees them as new tools automatically — you do not update the config or re-run the setup prompt to reach a newer model.
Use one RunAPI key inside WorkBuddy.
Add the MCP server once, pick any model from the catalog, and verify the connection before your next WorkBuddy task.