thiagogoncalves.com

What it is

Thiagogoncalves.com feels more like a public technical portfolio for Thiago Goncalves than a conventional SaaS website. The product line most relevant to developer tools is Continuity: a persistent memory and runtime governance tool for AI-assisted development. It has been published on the VS Code Marketplace, and the site mentions a paid Pro tier as well as paying users in production. The problem it addresses is that AI coding assistants often lose architectural decisions, constraints, and context across sessions, leading to inconsistent output.

Core capabilities

Continuity is designed around the idea of “context as infrastructure.” It uses bounded-retrieval middleware to control what gets injected into each round of interaction, with a default top-K of 3, a hard cap of 10, and a character limit per record. The article cites a test set of 2,189 decision records and claims roughly a 96.4% reduction in token injection compared with a static embedding baseline. On the retrieval side, it uses local Sentence-BERT, ONNX Runtime, and SQLite, emphasizing sub-15ms retrieval, on-device vector indexing, 100% local storage, and zero telemetry.

Security and integrations

The product’s most distinctive differentiator is security governance. It provides five layers of credential scrubbing: the MCP input boundary, AI output boundary, persistence write, persistence read, and MCP tool-result return. It also includes 27 provider-specific pattern categories plus a Shannon entropy fallback. Another key capability is the MCP tool interception layer, which performs Intercept, Pause, Evaluate, and Enforce before the LLM initiates file-system or external-service operations. For integrations, the article mentions the VS Code Extension API, CLI, MCP, and multiple LLM backends including Anthropic, Gemini, OpenAI, and Ollama. For self-hosting, it explicitly states that Ollama on TrueNAS SCALE has been used.

Pricing and documentation

Pricing information is limited. The page only says the product has evolved from bare metal to a paid Pro tier on the VS Code Marketplace and has active paying users; it does not provide specific pricing, free allowance, enterprise plan details, payment methods, or a refund policy. In terms of documentation quality, the captured content offers fairly in-depth explanations of the architecture, threat model, metrics, and technology stack, but lacks information aimed at regular users, such as an installation guide, configuration examples, FAQ, support SLA, and similar materials.

Pros, cons, and who it is for

Its strengths are clear security boundaries, a local-first approach, zero telemetry, replaceable multi-model backends, and a concrete engineering solution to credential contamination in AI memory. The drawbacks are that the public information still leans toward a personal résumé and technical showcase, while commercialization, support, editor coverage, and open-source strategy remain unclear. It is best suited to advanced developers or security-sensitive teams that make heavy use of AI coding assistants, need to preserve architectural decisions, and are concerned about credential leakage or tool misuse.

Access from China

The article does not provide information about network reachability, payment methods, or China-specific service availability, so its accessibility from China can only be considered unknown. If users need to access the VS Code Marketplace or backends related to OpenAI, Anthropic, or Gemini, the actual experience may be affected by network and account conditions. The local Ollama route may be more controllable for users in China. Comparable options include Cursor, Continue, Claude Code memory, Sourcegraph Cody, or a self-built Ollama + RAG setup.