Skip to content

Hydracept changelog

com.hydracept/vscode

4 changes recorded. MCP Harbor re-reads this listing from the official registry every six hours and asks the live server what it exposes, and keeps what changed between one look and the next.

  • llms.txt +17 −2

    https://hydracept.com/llms.txt

    − - Hydracept panel (PyPI hydracept >= 0.5.3): one MCP App ui://hydracept/app.html; seven states, not seven iframes− - Current published client: 0.5.3 (PyPI hydracept, npm @hydracept/sdk, NuGet Hydracept.Client)+ - Hydracept panel (PyPI hydracept >= 0.5.4): one MCP App ui://hydracept/app.html; seven states, not seven iframes+ - Current published client: 0.5.4 (PyPI hydracept, npm @hydracept/sdk, NuGet Hydracept.Client)+ - https://hydracept.com/case-studies+ ## Exact execution targets+ - Pin one provider and model: `"pin": {"target": "provider:model"}` on invoke or a durable job. Hydracept does not substitute; an unknown, ineligible, or unavailable target fails instead of running another model.+ - Pin support is limited to text.* capabilities today; other capability keys reject pin.target with 422 target_pin_unsupported, even when execution-targets lists models+ - Discover pinnable targets: GET /v1/capabilities/{key}/execution-targets (`id`, `eligible`, `availableNow`); check execution.supportsExplicitTarget first+ - The result reports `resolvedProvider` / `resolvedModel` / `servedCandidateId`; the receipt reports `route.routingPolicy: target_exact` and `targetRouting.substituted: false`+ - Docs: https://docs.hydracept.com/capabilities/+ ## Moderation and data classification+ - text.moderate.v1 classifies one string (`{"text": ...}`, <=20000 chars) as synchronous invoke only, free at the managed retail price+ - It returns evidence, not a verdict: `flagged` (may be null), provider-reported `categories`, `availability`. Your product owns the policy+ - moderation.content.v1 is the deprecated provider-aligned name for the same classifier+ - A declared context.dataClassification is a constraint, not authorization. ChildDirected is refused with 403 processing_not_permitted+ - Docs: https://docs.hydracept.com/moderation/+ - Docs: https://docs.hydracept.com/data-handling/+ - Docs: https://docs.hydracept.com/conversational-products/

  • llms.txt +4 −3

    https://hydracept.com/llms.txt

    − - Hydracept panel (PyPI hydracept >= 0.5.2): one MCP App ui://hydracept/app.html; seven states, not seven iframes− - Current published client: 0.5.2 (PyPI hydracept, npm @hydracept/sdk, NuGet Hydracept.Client)− Discovery includes readySummary, execution.managedAvailable, and output.kind.+ - Hydracept panel (PyPI hydracept >= 0.5.3): one MCP App ui://hydracept/app.html; seven states, not seven iframes+ - Current published client: 0.5.3 (PyPI hydracept, npm @hydracept/sdk, NuGet Hydracept.Client)+ Discovery includes readySummary, execution.managedInferenceAvailable, execution.trialEligible, funding.status, and output.kind.+ - https://hydracept.com/capabilities/moderation

  • llms.txt +11 −14

    https://hydracept.com/llms.txt

    − - In a git checkout: python -m hydracept init --apply --yes --json (binds stdio MCP to workspace secrets). Reload MCP once. Do not copy the workspace key into Plugins → Configure.− Free for individual developers. BYOK at 0% Hydracept fee.− Managed inference: PAYG wallet + 6% service fee where enabled.− Hydracept is an execution control plane for software and agents that need external capabilities — discovery, quoting, execution, durable jobs, recoverable history, artifacts, receipts, provenance, budgets, and provider abstraction. Games and media are important capability families, not the whole product. BYOK with 0% markup; managed inference is provider price + 6%. Durable text jobs on eligible models use deferred processing at 50% of standard token rates. One capability when you need it; a pro− ## Entity− Hydracept is:− - AI agent execution infrastructure− - external capabilities for AI agents− - durable AI jobs− - AI execution receipts− - BYOK capability execution− Workflow: An agent describes the work in plain language. Hydracept resolves a capability, quotes the spend, executes it synchronously or as a durable job, and returns artifacts with a receipt.− Install in Cursor or Claude Code: python -m hydracept init --apply --yes --json− Hosted MCP: https://api.hydracept.com/mcp+ - In a git checkout: python -m hydracept init --apply --yes --json (binds stdio MCP). Init may reuse an existing GitHub (`gh`) or Google Cloud (`gcloud`) user login, or HYDRACEPT_API_KEY, instead of opening a browser — see docs/authentication. Reload MCP once. Do not copy the workspace key into Plugins → Configure.+ Hydracept is AI agent execution infrastructure — one public API for external capabilities for AI agents, with durable AI jobs, AI execution receipts, and BYOK capability execution.+ Free plan ($0/mo) for individual developers; paid plans add capacity and org features.+ BYOK at 0% Hydracept fee.+ Managed inference: PAYG wallet + 6% Hydracept service fee on provider cost (no per-call minimum; billed at actual usage).+ Hydracept is a public API for software teams and coding agents to discover, quote, and run external AI capabilities (text, image, audio, and more) with durable jobs, artifacts, receipts, budgets, and BYOK or managed billing.+ Hydracept is a public API for discovery, quotes, execution, durable jobs, artifacts, receipts, and budgets. BYOK is 0% Hydracept fee; managed billing is provider cost plus 6%. Eligible durable text jobs can use deferred processing at 50% of standard token rates. Games and media are common examples, not the whole product. Hydracept is not a runtime game-engine inference plugin.+ ## Learn more+ Example workflow: An agent describes the work in plain language. Hydracept resolves a capability, quotes the spend, executes it synchronously or as a durable job, and returns artifacts with a receipt.+ Cursor/Claude plugin version (distributionVersion in plugin manifest) is separate from the PyPI/npm client (0.5.x).+ Hosted MCP: https://api.hydracept.com/mcp (compact default for new keys; use /mcp/full for all tools)

  • llms.txt +33 −13

    https://hydracept.com/llms.txt

    − - Hydracept panel (PyPI hydracept >= 0.3.9): one MCP App ui://hydracept/app.html; seven states, not seven iframes− Hydracept is an execution control plane for software and agents that need external capabilities, providing discovery, quoting, execution, durable jobs, recoverable history, artifacts, receipts, provenance, budgets, and provider abstraction.− - multi-provider AI generation API− - Unity AI generation− - reference-image generation− - transparent game assets− - cohesive sprite sheets− - agent MCP− Workflow: AI game assets from Cursor or Claude: reference image → cohesive sheet → transparent slices → Unity project, routed across providers with price and receipt provenance.− Canonical page: https://hydracept.com/agent-game-assets− - 5-seat early beta: US$25 covered managed inference, no card required. Tell your coding agent: install Hydracept (https://hydracept.com — python -m hydracept init --apply --yes --json) and sign in with GitHub, then email john@zencode.ca with your GitHub email or username for founder approval.− - Fallback application form: https://hydracept.com/beta− - https://hydracept.com/agent-game-assets+ - Hydracept panel (PyPI hydracept >= 0.5.2): one MCP App ui://hydracept/app.html; seven states, not seven iframes+ ## Client release+ - Current published client: 0.5.2 (PyPI hydracept, npm @hydracept/sdk, NuGet Hydracept.Client)+ - Migration guide: https://docs.hydracept.com/migrating-0.5.0/+ - Stream invoke SSE: GET /v1/executions/{executionId}/events (clients < 0.5.0 hit LEGACY_ROUTE_REMOVED on removed routes)+ - AI agent execution infrastructure+ - external capabilities for AI agents+ - durable AI jobs+ - AI execution receipts+ - BYOK capability execution+ Workflow: An agent describes the work in plain language. Hydracept resolves a capability, quotes the spend, executes it synchronously or as a durable job, and returns artifacts with a receipt.+ Canonical page: https://hydracept.com/agent-execution-infrastructure+ - A yes-or-no decision over structured state+ - Self-serve sign-in (GitHub or Google) at /start — no credit card. Agents: python -m hydracept init --apply --yes --json.+ - Comped managed credit (limited): email support@hydracept.com after sign-in.+ - Optional beta application form: https://hydracept.com/beta+ - https://hydracept.com/agent-execution-infrastructure+ - https://hydracept.com/receipts+ - https://hydracept.com/agent-game-assets+ - https://hydracept.com/changelog+ - https://hydracept.com/trust+ - https://hydracept.com/license+ - https://hydracept.com/capabilities/image-generation+ - https://hydracept.com/capabilities/video-generation+ - https://hydracept.com/capabilities/voice+ - https://hydracept.com/capabilities/music+ - https://hydracept.com/capabilities/sound-effects+ - https://hydracept.com/capabilities/3d+ - https://hydracept.com/capabilities/domains+ - https://hydracept.com/capabilities/translation+ - https://hydracept.com/capabilities/text-structured+ - https://hydracept.com/capabilities/text-reasoning+ - https://hydracept.com/capabilities/compute