Start from a seed, not a blank prompt.
SeedSpec turns product and domain knowledge into a structured starting point for AI agents—with tools to author it well and a package format built to travel.
What’s in a SeedSpec?
A SeedSpec is a portable package for what should be understood before implementation begins: the intended outcome, boundaries, meaningful choices, relevant guidance, and how success should be evaluated.
The name describes its role. It is a starting point meant to grow into a particular result—not a complete implementation and not a promise that two agents will produce identical software.
Why SeedSpec
A specification is useful before it becomes a package. SeedSpec adds value at three points: improving the starting state, helping the author develop the material, and making the result cleanly distributable.
Better starting point
Start with more than a request: product intent, constraints, open decisions, relevant context, and success criteria already in view. The agent spends less of the implementation inventing the product.
Better authoring
Begin in Markdown, then use guided review to surface ambiguity, missing policy, adopter choices, provenance, and evidence. The tools help an author make the specification fuller; they do not pretend every author did.
Clean distribution
Package human-readable intent with stable semantic roles, versions, and exact digests. Share it through a team library, vendor catalog, public collection, or marketplace without losing which material means what.
What we know about specifications
The most important result from our evaluations is simple: explicit intent gives an agent a better opportunity to preserve important product decisions than leaving those decisions implicit.
Good Markdown can already be excellent. SeedSpec’s opportunity is not “more files means a better spec.” It is guided authoring, semantic ownership, and a package that can be reviewed, adapted, and distributed without every recipient reconstructing what each passage means.
- Product intent
- What must be true.
- Configuration
- What the adopter may choose.
- Decisions and provenance
- Who selected what, and why.
- Acceptance
- How success should be evaluated.
- Implementation resources
- Guidance about how to do the work.
- Evidence
- What has actually been observed or verified.
Try Family Hub
Family Hub packages the product decisions behind a household coordination app: roles, privacy, recurring work, optional allowance and meal-planning behavior, and observable acceptance.
npx --yes @seedspec/cli@next begin https://github.com/SeedSpec/reference-solutions/tree/main/solutions/family-hub/seedspecbegin acquires the package, validates it, and presents the handoff without executing package content.
See how it feels to use
These walkthroughs connect the authored specification to adopter choices, implementation paths, and the output an agent should produce. Where no realization has been committed, the pages say so directly.
Operations Dashboard Starter Kit
Turn live operations into focused action.A reusable operations console with trustworthy metrics, prioritized records, saved views, governed actions, and safe exports.
- Configured vocabulary
- Metric freshness
- Role-aware actions
- Audit history
See how the package becomes a Returns Operations console through configuration, then how an agent would map it to the real environment.
Customer Onboarding Orchestrator
Make every handoff owned and visible.Turn an approved deal into milestones, owners, customer updates, stalled-work escalation, and auditable completion.
- Three implementation paths
- Milestone ownership
- Escalation
- Evidence
Follow one durable outcome through configured SaaS, an automation platform, or a dedicated coordination service.
Use cases
Different people can create, distribute, approve, adapt, and implement specifications. The package remains open and inspectable while each organization adds its own trust, support, governance, or commercial model.
Internal enterprise libraries
Publish approved product, feature, workflow, and configuration packages with company terminology, policy, guidance, and release expectations.
Explore this use case →Vendor-produced specifications
Package the best way to build with a platform—including its data model, APIs, extension points, examples, and operational guidance.
Explore this use case →Consultancies and agencies
Reuse proven product and domain thinking while adapting every implementation to the customer’s stack, policies, and operating model.
Explore this use case →Libraries and marketplaces
Distribute inspectable solution knowledge with discovery, curation, reviews, reputation, licensing, and support around an open package format.
Explore this use case →Built deliberately, in public
SeedSpec is more than a format sketch. The project includes guided authoring workflows, published protocol, runtime, and CLI packages, a conformance suite, reference specifications, implementation resources, and an open evaluation program.
The protocol is precise because tools need dependable interoperability. The rest of the system exists to help people author better starting points, distribute them cleanly, and give agents more useful context.
See the work on GitHub ↗