Integration

EcoHash integration

EcoHash (EcoLink) runs open models behind an OpenAI-compatible inference API. An agent can chat with text and vision models (stateless chat completions or stateful multi-turn responses), create embeddings, rerank documents, generate and edit images, synthesize, transcribe, translate and diarize speech, clone a voice or edit recorded speech, generate music, generate videos (with cost quotes and reference media uploads), read the live model catalog and prices, and read the account balance, transactions and this key's usage. Every inference call is billed to the account's prepaid credit balance (per token, image, second of audio or video usage). Left out: GPU compute (instances, clusters, reservations, storage, SSH keys, container images, templates, user inference instances), streaming answers, legacy text completions and score (their bodies are undocumented) and the OpenAI-shaped video aliases. Authentication is an API key (it starts with eco_) from the EcoHash console's API Keys page, sent as a Bearer token; the balance and transaction tools also need Compute access (read-only is enough) on the key.

Connect EcoHash How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in EcoHash.Built from the 33 EcoHash tools below, on the connection you add once.

What you can do with EcoHash

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.

Response, 3 operations

cancel_responseCancel responsePOSTAgents: ask firstCancels a response created by create_response. Only the account that created it can cancel it. The vendor documents no answer body; it is returned as sent.
create_responseCreate responsePOSTAgents: ask firstRuns one turn of a stateful conversation (text platform models only): send the new input and the previous response id; EcoHash keeps the earlier turns for 24 hours. Billed per token.
get_responseGet responseGETAgents: ask firstReturns a response created by create_response (same shape as the create answer). Only the account that created it can read it.

Image, 2 operations

edit_imageEdit imagePOSTAgents: ask firstEdits a stored image from a prompt (optionally with a mask) with an edit-capable model; answers like generate_image (data[].b64_json). Billed per image.
generate_imageGenerate imagePOSTAgents: ask firstGenerates one image from a text prompt (OpenAI-compatible) and returns it as base64 or as a link valid 24 hours (response_format). Billed per image.

Platform model, 2 operations

get_platform_modelGet platform modelGETAgents: ask firstReturns one public platform model by its model_id: category, modalities, limits, regions and live prices.
list_platform_modelsList platform modelsGETAgents: ask firstLists every public platform model with its model_id, category, modalities, context length, regions and live prices. No key needed.

Reference media, 2 operations

refresh_reference_media_urlRefresh reference media urlGETAgents: ask firstReturns a fresh time-limited url for reference media uploaded earlier with upload_reference_media.
upload_reference_mediaUpload reference mediaPOSTAgents: ask firstUploads a stored image, video or audio file (up to 200 MB) as reference media for generate_video. Returns a url valid 30 minutes and an id to refresh it.

Speech, 2 operations

create_speechCreate speechPOSTAgents: ask firstSynthesizes speech from text (up to 4000 characters) with a text-to-speech voice and returns the audio file (mp3, wav or opus). Billed per second of audio.
edit_speechEdit speechPOSTAgents: ask firstEdits recorded speech by its transcript: send the audio, its original and edited transcripts; only the changed words are regenerated in the same voice. Returns the audio file. Billed as speech.

Video job, 2 operations

get_video_jobGet video jobGETAgents: ask firstReturns a video job's status and, once completed, a download url valid 30 minutes (each read issues a fresh one); a failed job carries code, error and retryable.
list_video_jobsList video jobsGETAgents: ask firstLists the account's video generation jobs, newest first, with status, size, billing and failure codes. Page with next_cursor.

Audio translation, 1 operation

translate_audioTranslate audioPOSTAgents: ask firstTranscribes a stored audio file in any supported language and returns the text translated into English (OpenAI-compatible). Not every speech-to-text model translates. Billed like transcription.

Balance, 1 operation

get_balanceGet balanceGETAgents: ask firstReturns the account's credit balance and post-paid state (pending usage, warning and overdue dates). The key needs Compute access (read-only is enough).

Chat completion, 1 operation

create_chat_completionCreate chat completionPOSTAgents: ask firstGenerates the next assistant message for a conversation with a text or vision model (platform model or your own inference instance). Returns choices[0].message and token usage. Billed per input and output token.

Cloned speech, 1 operation

clone_voice_speechClone voice speechPOSTAgents: ask firstSpeaks the input text in the voice of a short reference clip sent with the request and returns the audio file. Use only a voice you have the right to clone. Billed as speech.

Document ranking, 1 operation

rerank_documentsRerank documentsPOSTAgents: ask firstScores candidate documents against a query with a cross-encoder (Cohere-style rerank) and returns them best first, with each one's original index. Billed per token.

Embedding, 1 operation

create_embeddingsCreate embeddingsPOSTAgents: ask firstTurns texts into embedding vectors (OpenAI-compatible) for search, RAG or clustering; one vector per input text, in order. Billed per input token.

Health, 1 operation

get_healthGet healthGETAgents: ask firstTells whether the EcoHash API is running (status ok).

Model, 1 operation

list_modelsList modelsGETAgents: ask firstLists the text models this key can call in the OpenAI model-list format, including your running inference instances. Video and other models are in list_platform_models.

Model pricing, 1 operation

get_model_pricingGet model pricingGETAgents: ask firstReturns one model's live prices: per-token, per-image or per-second-of-audio rates, and for a video model its per-band rates (video_tiers).

Music, 1 operation

generate_musicGenerate musicPOSTAgents: ask firstGenerates music or a sound effect from a text prompt and returns the audio file (WAV). Billed.

Region, 1 operation

list_regionsList regionsGETAgents: ask firstLists the active EcoHash regions: id, name, country, GPU type and status.

Speaker diarization, 1 operation

diarize_audioDiarize audioPOSTAgents: ask firstDetects who spoke when in a stored audio file and returns speaker turns with start and end times; it does not transcribe. The answer is returned as sent. Billed.

Transaction, 1 operation

list_transactionsList transactionsGETAgents: ask firstLists the account's credit transactions (top-ups, usage, refunds, bonuses) newest first: billing data, card country of top-ups included. Account owners only; the key needs Compute access (read-only).

Transcription, 1 operation

transcribe_audioTranscribe audioPOSTAgents: ask firstTranscribes a stored audio file (at most 25 MB and 30 minutes) to text (OpenAI-compatible). verbose_json adds language, duration and timed segments. Billed per second of audio.

Tts voice, 1 operation

list_tts_voicesList tts voicesGETAgents: ask firstLists the voices of each deployed text-to-speech model. Where a model is not multilingual the voice selects the language: pick one whose lang matches the text.

Usage record, 1 operation

list_usageList usageGETAgents: ask firstLists the calling API key's own usage records (model, tokens, cost, latency, status), newest first; never other keys or the account total. Defaults to the last month.

Video, 1 operation

generate_videoGenerate videoPOSTAgents: ask firstGenerates a video from a prompt, optionally guided by reference image, video or audio URLs. Billed on real usage (resolution, duration); quote_video estimates it. Waits for the job and returns the finished result; gives up after 10 minutes.

Video file, 1 operation

download_videoDownload videoGETAgents: ask firstDownloads the finished video of a completed job as a stored file. A job still generating answers 409.

Video price, 1 operation

list_video_pricingList video pricingGETAgents: ask firstLists the public video price matrix: for each active video model, the price per resolution band, with and without a reference video.

Video quote, 1 operation

quote_videoQuote videoPOSTAgents: ask firstEstimates the pricing band and cost range of a generate_video request (same fields) without generating anything. Nothing is billed.

Use EcoHash from Claude Code, Cursor or Codex

Connect EcoHash in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its EcoHash 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 EcoHash

Create your workspace
Sign up free. Nothing to install.
Connect EcoHash once
Add your EcoHash 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.