Tiden

Product Introduction

1. Definition: Tiden is an intent verification and specification management platform for software development teams using AI coding agents. It operates as a specialized layer in the CI/CD pipeline that automatically captures, versions, and validates product requirements ("specs") against code changes.
2. Core Value Proposition: Tiden exists to solve the verification gap created by AI coding agents like Claude Code and Cursor. While these agents accelerate code generation, they introduce risk by potentially misinterpreting intent. Tiden's core value is providing deterministic, test-based verification that agent-written code fulfills the original business requirements, thereby enabling faster and more reliable shipping of AI-generated features.

Main Features

1. Automated Spec Capture & Mapping: Tiden statically analyzes codebases across multiple languages (e.g., TypeScript, Go) and integrates with issue trackers (Jira, Linear) to automatically generate and map a living specification of what the product does. It identifies gaps where functionality lacks corresponding tests, creating a real-time "product snapshot."
2. Session-Based Intent Verification: During an AI coding agent session, Tiden locks the session against a set of actionable "promises" derived from specs. The session cannot be closed until all promises are proven by passing tests run within the agent's own environment, ensuring agreement is not just internal to the AI model.
3. Versioned Specs & Branch-Aware Diffing: Specifications are stored in a tiden/ directory and version-controlled alongside code using Git. Tiden automatically drafts updates to these specs on every merge, providing a clear diff of "what changed" in the code versus "what it was supposed to do," even for changes made outside of its sessions.
4. Release-Centric Risk Dashboard: Tiden aggregates verification results per release, showing component-level pass/fail status and "risk" scores based on untested specs. This creates an audit trail where production errors can be traced directly back to the specific release and untested promise that caused them.

Problems Solved

1. Pain Point: The disconnect between AI coding agent output and actual user intent. Manual verification of agent-written code is slow, error-prone, and negates the speed benefit of using agents. Traditional CI/CD pipelines test for regressions but cannot verify if the implemented logic matches the original requirement.
2. Target Audience: Engineering leads and development teams in mid-to-large size companies who are actively deploying AI coding agents (Claude Code, Cursor, GitHub Copilot) into their daily workflow and need to maintain velocity without sacrificing code quality or requirement fidelity. DevOps engineers seeking to integrate agent verification into pipelines.
3. Use Cases: * Governing feature development by AI agents against product requirement documents (PRDs). * Automating compliance and audit trails for changes made by AI. * Onboarding new engineers by providing a verified, living map of system behavior. * Refactoring or extending legacy systems with AI assistance while ensuring existing behavior is preserved and documented.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):

   • Pros: Creates a single source of truth for product behavior. Shifts verification left by integrating directly into the coding session. Provides objective, test-driven verdicts, not model-generated "grades." Excellent integration with existing developer tools (Git, Jira, GitHub).
   • Cons: Requires a test suite to be effective; it manages and enforces tests but does not generate them autonomously. Adds a new layer of abstraction (specs) to maintain. Primarily valuable for teams already heavily invested in AI coding agents; less critical for traditional manual development.

2. Key Alternatives & Differentiation:

   • Traditional Unit/Integration Testing (Jest, pytest): These verify code works but not that it matches a centralized, versioned requirement. Tiden differentiates by generating and managing the specification that tests must prove, linking them directly to product intent.
   • AI-Powered Code Review (PullRequest.ai, CodiumAI): These tools analyze code diffs for quality and suggest tests. Tiden is more prescriptive and integrated, governing the entire development session from intent to verified merge, rather than acting as a post-hoc reviewer.
   • Behavior-Driven Development (BDD) Frameworks (Cucumber): Like BDD, Tiden uses plain-language specs. However, Tiden's specs are automatically inferred and updated from code and issues, whereas BDD specs are typically written manually upfront, creating maintenance overhead.

Frequently Asked Questions (FAQ)

1. How does Tiden verify AI-generated code differently from a human code review? Tiden provides deterministic verification by running automated tests against versioned specifications during the AI coding session. Unlike human review, which is subjective and asynchronous, Tiden offers immediate, objective feedback based on the agreed-upon contract, preventing the merge of code that doesn't fulfill its promises.
2. Does Tiden work with any AI coding agent or only specific ones? Tiden is designed to work alongside popular AI coding agents like Claude Code, Cursor, GitHub Copilot, and Codex. It operates by managing the session and verification environment, making it largely agent-agnostic for agents that can operate within its configured workflow.
3. What happens if my existing test coverage is low? Tiden will map your code and highlight all specs (functionalities) that lack test coverage as "risk." You can then use its integration to task your AI agent with writing the missing tests, thereby incrementally improving coverage while verifying new work.
4. Is Tiden a replacement for my CI/CD pipeline or test suite? No, Tiden is a complementary layer. It leverages your existing CI/CD pipeline and test suite to perform its verification. It adds the critical capability of intent management and ensures tests are aligned with a master record of what the system is supposed to do.
5. How does Tiden handle changes made directly by developers without using an AI agent? Tiden monitors the Git repository. When a merge occurs (agent or human), it automatically analyzes the diff and drafts updates to the affected specifications, ensuring the living documentation stays synchronized with the actual codebase.