blisspixel/primr

Primr

Point it at a company's website. It reads their DNS records, their job postings, and fifty pages of their site, then drafts a strategic brief: what they appear to be building, where they look constrained, and which questions are worth asking.

Summaries of published articles are a commodity; any chat assistant's deep-research mode produces one. Primr is built for the layer underneath: primary-source signals that aren't in the articles. DNS records reveal the real tech stack. Job postings reveal what a company is actually building right now. The site corpus, external research, and that signal layer get triangulated into a long-form strategic analysis covering competitive positioning, likely economics, operating constraints, and working hypotheses, with confidence labels distinguishing what's confirmed, what's reported, and what's inference. The labels are model-applied and spot-audited for traceability (primr calibrate); treat them as editorial discipline, not ground truth.

primr "ExampleCo" https://example.co

About 23-35 minutes later: a 23-section strategic analysis as Markdown and DOCX, with dense references consolidated at the end. ~$0.79 in API costs when both GEMINI_API_KEY and XAI_API_KEY are set (Grok 4.3 for reasoning with cached input, Gemini 3.1 Flash-Lite for bulk writing; the v1.24.0 default after a cross-provider eval). XAI-only setups stay on the legacy Grok-NR writing path at ~$4.27/run.

Primr is local-first and CLI-first; it also runs as an MCP server and a Claude Skill so agents can drive it (details below).

Not a developer? Y