SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsAI codingMySpec

MySpec

MySpec is a specification generation tool for AI programming scenarios, designed to solve the code hallucination problem that arises when AI coding agents such as Cursor, Claude Code, and Windsurf lack clear specifications. Through a structured interview process, it transforms vague ideas into a deterministic 4-file specification package (constitution.md, requirements.md, solution, tasks) for AI coding agents to follow. The core philosophy is "clarity precedes code," replacing vague chat prompts with immutable architectural contracts to reduce problems such as agents forgetting non-goals, inventing APIs, and overriding architecture. The tool provides a four-step workflow from description, interview, and specification to export, supports Markdown and Mermaid editing, ZIP export, and can integrate with IDE agents through an MCP server. It offers three plans: Free, Pro, and Team. The free version includes 20 projects and 100 specification files.

Open website

About this tool

MySpec is a specification (spec) generation tool positioned at the front end of AI-assisted software development, mainly addressing the "code hallucination" problem that arises when AI coding agents lack clear, structured instructions. It does not directly write code. Instead, before code generation, through an interview in the style of a senior AI architect, it helps users transform vague ideas into a deterministic four-file specification package for AI coding agents such as Cursor, Claude Code, and Windsurf to execute.

Purpose and Core Philosophy MySpec's core propositions are "CLARITY PRECEDES CODE" and "SPEC DRIVEN IS DETERMINISTIC." It targets the problem that a purely prompt-driven development approach can easily cause agents, after multiple rounds of conversation, to forget non-goals, invent nonexistent APIs, and override existing architecture. MySpec replaces vague chat prompts with an immutable four-file architectural contract, keeping rules immutable across all agent prompts and avoiding context degradation.

Main Capabilities

  1. Four-file specification package: generates constitution.md (constitution/non-negotiable rules), requirements.md (requirements), solution (solution), and tasks (tasks) as the basis for AI coding agents to build from. In the materials, constitution.md and requirements.md are explicitly listed file names, while solution and tasks are not given the .md extension.
  2. Senior AI architect interview: uses a structured interview to uncover edge cases and transform messy ideas into clear files.
  3. Deterministic contract: replaces vague prompts with immutable specifications to reduce agents deviating from constraints.
  4. Four-step workflow: DESCRIBE -> INTERVIEW -> SPECIFY -> EXPORT, with no blank pages and no guessing.
  5. Editing and export: provides a Markdown + Mermaid editor and supports ZIP export of the specification package.
  6. MCP integration: provides the HTTP MCP server mcp.myspec.dev, aimed at hosted agents, public share links, and custom skills; the materials also mention @myspec/mcp-server integration, which can allow IDE agents to pull structured context through stdio.
  7. Compatible workflows: supports Spec Kit-compatible greenfield workflow (spec.md, plan.md, tasks.md); supports brownfield workflow, including MySpec proposals + requirement deltas, as well as OpenSpec-compatible changes.
  8. Team features: organizations, invitations, personal API tokens (for unattended MCP), shared projects and roles, centralized billing.
  9. Public share links: spec share links can be generated.

How to use According to the material, the typical process is: the user first describes a messy idea, MySpec conducts a structured interview, then generates four refined files, and finally they can be exported. Users can write specs in the browser, export to Git, and IDE agents immediately pull structured context via stdio. Specific installation commands, API call details, and similar information are not provided.

Use cases Suitable for software architects, engineering leads, and technical founders using AI coding agents; suitable for teams that want to constrain AI agents with deterministic specs, reduce PR review time, and avoid production defects caused by undeclared assumptions; supports greenfield and brownfield projects. User feedback cited in the material mentions that PR review time dropped by 65% because PRs reference precise requirement IDs; acceptance criteria in EARS+ format provide distributed teams with a single source of truth shared by humans and AI agents.

Plans and limitations The Free plan covers 20 projects, 100 spec files, 20 public share links per month, a Markdown + Mermaid editor, spec package ZIP export, and community support. The Pro plan includes all Free plan features, 100 projects, unlimited specs, 5x AI usage limits, advanced integrations, GitHub and VS Code (available after launch), and priority email support. The Team plan includes all Pro plan features, unlimited projects and specs, 10x AI usage limits per seat (minimum 2 seats), team collaboration, shared projects and roles, and centralized billing. It is currently in Open Beta, offers a free tier, and does not require a credit card. GitHub and VS Code integrations are marked as "when shipped," meaning they are not yet live. Information such as pricing amounts, specific models, privacy policy, and data storage locations is not provided.

Related tools

Browse category