Hopsule
Hopsule provides a persistent decision and memory layer for AI-assisted software development, capturing architectural decisions and project context so AI coding tools maintain consistent knowledge across sessions. The platform structures decisions with a formal lifecycle, creates searchable memories, and surfaces advisory enforcement warnings directly in IDEs, CLI, and CI/CD pipelines without ever blocking workflows or storing source code.
- Artificial Intelligence
- AI Agents
- Developer Tools
- Software Only
Funding
Founders
Product
Problem
AI coding assistants operate without persistent context, forcing engineering teams to re-explain architectural decisions in every session and leading to lost institutional knowledge, repeated mistakes, and silent architectural drift. Decisions scattered across Slack threads, docs, and individual memories create inconsistencies that compound as teams grow and new members onboard.
Solution
Hopsule provides a persistent memory layer for AI-assisted development, capturing architectural decisions and project memories as structured, versioned primitives that every AI tool can access. The platform gives decisions a formal lifecycle—Draft, Pending, Accepted, Deprecated—where accepted decisions become immutable and serve as the system of record for the entire team. Memories store conventions, migration notes, and tech stack context without a lifecycle, remaining as background knowledge for AI assistants. Capsules bundle decisions and memories into portable, shareable context packs that the MCP server materializes into structured prompts, with freeze and supersede capabilities for versioned snapshots. The knowledge graph maps relationships between decisions, memories, tasks, and capsules with typed connections, while Hopper, a RAG-based chat assistant, grounds every response in the project's actual architecture. Enforcement is advisory-only, surfacing warnings in the IDE, PRs, and CI/CD without ever blocking merges or deployments.
Target Audience
Primary customers are software engineering teams—from solo developers to large organizations—using AI coding assistants like Cursor, Claude, Copilot, and Windsurf who need persistent architectural context and governance across their development workflow.
Features
- Decision lifecycle management with immutable accepted decisions and deprecation workflow, including statement, rationale, scope key, affected file paths, and tags
- Persistent memories with no lifecycle, linkable to decisions, scoped by file paths, tagged for search, and source status tracking showing active, partially orphaned, or fully orphaned code
- Capsules as curated context packs holding references (not copies) with freeze and supersede capabilities for immutable snapshots
- Interactive knowledge graph with typed connections—explicit, semantic, temporal, causal—for visualizing architecture relationships and identifying orphan decisions
- Hopper RAG-based chat assistant that searches decisions, memories, and tasks before answering, supporting streaming, image uploads, conversation compression, and multilingual responses
- Advisory-only enforcement with three severity levels (HIGH, MEDIUM, LOW) across IDE on save, GitHub App in PRs, and CLI in CI/CD, with deterministic conflict detection based on scope keys
- Taskify integration creating tasks directly from the IDE, with feedback loops that validate or challenge decisions and track decision health
- MCP server and CLI tool for syncing context from the terminal, plus IDE extension providing advisory warnings inline