Context Portal MCP (ConPort)

Context Portal, or ConPort, serves as a dedicated memory bank for software projects that need to be understood by AI assistants. By replacing ad‑hoc text files with a structured, database‑backed knowledge base, it allows developers to capture decisions, architectural patterns, glossaries, and progress notes in a way that AI can query efficiently. The result is a project‑specific knowledge graph enriched with vector embeddings, enabling semantic search and Retrieval Augmented Generation (RAG) so that assistants can return precise, up‑to‑date answers tailored to the current codebase.

ConPort’s core value lies in its structured storage. Each workspace gets its own SQLite database that automatically records context items—such as design decisions, task status, or component specifications—and the explicit relationships between them. This structure removes ambiguity that plagues plain‑text memories and gives AI a clear map of how concepts relate, which improves both accuracy and explainability. The embedded vectors support fuzzy matching, so a query like “what is the authentication flow?” will surface all relevant entries even if phrased differently.

The server exposes a rich set of MCP tools. Developers can add, update, or delete context entries via simple API calls; query the graph with natural language prompts that are translated into structured requests; and retrieve embeddings for RAG pipelines. Because ConPort is built on FastAPI, it can run in STDIO mode for tight IDE integration or as a standalone HTTP service for broader tooling. Multi‑workspace support is built in through the parameter, making it ideal for monorepos or teams working on multiple projects simultaneously.

In practice, ConPort shines in scenarios where AI assistants need to reference architectural decisions or policy constraints that evolve over time. For example, a developer asking the assistant to refactor a module will receive not only code suggestions but also links to related decisions and constraints stored in ConPort. Similarly, onboarding new team members can be accelerated by querying the knowledge graph for glossaries or design rationales, rather than hunting through documentation. The database’s migration system (Alembic) guarantees that schema changes are applied safely, so teams can extend the context model without losing data.

Overall, ConPort offers a robust, queryable backend that turns scattered project notes into actionable AI knowledge. By integrating seamlessly with MCP‑compatible workflows, it empowers developers to build smarter, context‑aware assistants that truly understand the intricacies of their codebases.