About Hydracept
Hydracept is a hosted execution control plane for software and AI agents that need external capabilities such as image generation, reasoning, and compute. It supports capability discovery, durable jobs, artifacts, receipts, budgets, and BYOK or managed billing.
Key features
- 01Live capability discovery
- 02Durable capability jobs
- 03Artifact and receipt tracking
- 04Budgets and BYOK
- 05Hosted and stdio MCP
Use cases
- 01Find an image generation capability
- 02Submit and track external jobs
- 03Use BYOK for agent workloads
MCP support
Hydracept exposes tools through a hosted MCP endpoint or a stdio MCP server. Agents can use it to discover capabilities and run jobs.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://api.hydracept.com/mcpclaude mcp add --transport http com-hydracept-hydracept https://api.hydracept.com/mcp{"mcpServers":{"com-hydracept-hydracept":{"type":"http","url":"https://api.hydracept.com/mcp"}}}Open in Cursor{"servers":{"com-hydracept-hydracept":{"type":"http","url":"https://api.hydracept.com/mcp"}}}Open in VS Code{"mcpServers":{"com-hydracept-hydracept":{"type":"http","url":"https://api.hydracept.com/mcp"}}}{"mcpServers":{"com-hydracept-hydracept":{"serverUrl":"https://api.hydracept.com/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
https://api.hydracept.com/mcpAuth requiredTools
9| Parameter | Type | Required | Description |
|---|---|---|---|
| refresh | boolean | No | Stdio: when true, also verify session over the network. Ignored on hosted MCP. |
Details
From the project READMEHydracept
Hydracept is an execution control plane for software and AI agents that need external capabilities—text, reasoning, image, audio, video, 3D, domains, and compute—through one API with discovery, quotes, durable jobs, artifacts, receipts, budgets, and BYOK. Games and media are common examples; a headline workflow is AI game assets from Cursor or Claude Code into Unity (workflow overview).
Pricing (summary): Free plan ($0/mo) for individuals; paid plans add capacity. BYOK at 0% Hydracept fee; managed inference is provider cost + 6% service fee (actual usage, no per-call minimum). Eligible durable text jobs can use deferred processing at 50% of standard token rates. Details: hydracept.com/pricing and Billing docs.
Install
| Surface | Command |
|---|---|
| Python CLI & SDK | pip install -U hydracept |
| TypeScript | npm install @hydracept/sdk |
| .NET | dotnet add package Hydracept.Client |
| Unity (UPM git) | com.hydracept.unity via Git URL — see Unity integration (OpenUPM listing pending owner submit) |
Public HTTP API: https://api.hydracept.com · OpenAPI: hydracept-v1.json · Agent context: GET https://api.hydracept.com/v1/agent-context
Bootstrap
pip install -U hydracept
python -m hydracept init --apply --yes --json
python -m hydracept doctor
Studio (browser): app.hydracept.com/login. For CI or headless use, set HYDRACEPT_API_KEY and run the same init command with --ci when you need setup grants. Workspace files live under .hydracept/ — see Authentication.
Minimal example
One-shot image generation (composes init when the workspace can be bootstrapped):
python -m hydracept run image.generate.v1 --prompt "transparent 1024x1024 blue slime icon" --json
Python SDK (after init binds the workspace):
from hydracept.client import HydraceptClient
client = HydraceptClient.from_workspace()
job = client.submit_capability_job(
"image.generate.v1",
{
"input": {"prompt": "cute slime icon, flat game art"},
"execution": {"executionPreference": "automatic"},
"idempotencyKey": "demo-1",
},
)
finished = client.wait_for_job(job["jobId"])
receipt = client.get_job_receipt(job["jobId"])
Discover capabilities live (do not hard-code keys from this README):
python -m hydracept capabilities find "generate an image"
curl -sS -H "Authorization: Bearer $HYDRACEPT_API_KEY" https://api.hydracept.com/v1/capabilities
More examples: public/examples/.
MCP (Cursor & Claude Code)
In a project checkout, stdio MCP is the default after init:
python -m hydracept init --apply --yes --json # binds MCP to workspace secrets; reload MCP if prompted
python -m hydracept mcp serve # stdio server
python -m hydracept agents install --auto # Cursor / Claude Code plugin + hooks
- Plugin & install links: hydracept.com/plugin
- Hosted MCP (no repo checkout):
https://api.hydracept.com/mcpwith bearerHYDRACEPT_API_KEY - MCP Registry name:
com.hydracept/hydracept
Do not paste workspace API keys into IDE plugin config when stdio bind is available. Full agent guide: Coding Agents.
Documentation
| Topic | Link |
|---|---|
| Docs home | docs.hydracept.com |
| Quick start (detailed) | Quick Start |
| Capabilities & jobs | Capabilities, Jobs & receipts |
| Image production | Image production |
| Unity | Unity integration |
| BYOK & billing | Connections / BYOK, Billing |
| LLM-oriented summary | hydracept.com/llms.txt |
| Product manifest | /.well-known/hydracept.json |
Repository layout
clients/python— PyPI packagehydracept(CLI, SDK, local MCP)clients/typescript— npm@hydracept/sdkclients/csharp— NuGetHydracept.Clientpackages/unity— Unity Editor packagepublic/docs— source for docs.hydracept.com (Quick Start content; this root README is the GitHub front door)
License
MIT — see LICENSE. Package metadata on PyPI, npm, and NuGet also lists MIT.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.