Skip to content

Getting Started

Dong Dai edited this page May 26, 2026 · 2 revisions

This guide walks you from install to your first successful chat. If you hit a snag, check the FAQ or open a Discussion.

1. Install

Option A — npm (recommended)

npm install -g research-copilot
research-copilot

2. Sign in

Research Copilot picks the cheapest working auth method it finds, in this order:

ChatGPT subscription → Claude subscription → OpenAI API key → Anthropic API key

Easiest: use a subscription (no API key, no metered billing)

  1. Open the model selector at the top of the chat pane.
  2. Pick a GPT-5.4 (sub) entry (ChatGPT Pro / Plus) or a Claude … (sub) entry (Claude Pro / Max).
  3. Complete the OAuth flow in your browser.

Credentials are stored in the OS keychain and refreshed automatically.

Alternative: paste an API key

  1. Open the unified settings panel with Cmd + . (or click the gear icon).
  2. Go to the API Keys tab.
  3. Paste your key into ANTHROPIC_API_KEY or OPENAI_API_KEY.
  4. Click Save & Continue.

Keys are written to ~/.research-copilot/config.json.

Tip: at least one of Anthropic / OpenAI (API key or ChatGPT subscription) is required. The rest are optional.

Optional supporting keys

Key What it enables Without it
BRAVE_API_KEY General web search via Brave Falls back to arXiv-only academic search
OPENROUTER_API_KEY AI-generated scientific diagrams (scientific-schematics skill) Other skills still work

Semantic Scholar, arXiv, OpenAlex, and DBLP work without any key.

3. Your first chat

  1. Pick or create a workspace folder. Research Copilot stores artifacts and memory under <workspace>/.research-pilot/.
  2. Type a request in the chat pane. A few starting points:
    • "Find recent papers on quantum error correction and save the top 5 as notes."
    • "Load this CSV and plot a correlation heatmap." (attach a file via the paperclip)
    • "Draft the introduction for a NeurIPS-style paper on ."
  3. The agent will use tools — literature search, web fetch, Python analysis, file edits — as needed. Tool calls appear in the chat so you can watch what it's doing.

4. Turn on Paper Wiki (optional but recommended)

Paper Wiki is a background agent that indexes every paper you touch into a cross-project knowledge base.

  • Open settings (Cmd + .) → Paper Wiki tab.
  • Pick a model you're comfortable paying for (subscription-backed is cheapest; "Auto" follows the system priority).
  • Pick a speed preset.

Expect roughly 8K–25K input / 2K–4K output tokens per paper. It's disabled by default.

5. Add custom skills (optional)

Drop a Markdown file at <workspace>/.pi/skills/<name>/SKILL.md:

---
id: my-skill
name: My Skill
shortDescription: Brief description of what this skill does
---

Summary loaded at startup.

## Procedures
Detailed guidance loaded on demand when the skill is activated.

Skills are auto-discovered from three locations (later overrides earlier):

  1. Shipped builtin skills
  2. ~/.research-pilot/skills/ — user-global
  3. <workspace>/.pi/skills/ — project-specific

Next steps

Clone this wiki locally