> ## Documentation Index > Fetch the complete documentation index at: https://docs.valyu.ai/llms.txt > Use this file to discover all available pages before exploring further. # Use Valyu in Claude or ChatGPT > Add Valyu as a custom connector in Claude or ChatGPT with no code, so your assistant can search SEC filings, papers, clinical trials and the web with citations. Your assistant is only as good as what it can look up. Connect Valyu and Claude or ChatGPT can search SEC filings, scientific papers, clinical trials, patents, market data and the web, and its answers can cite the sources they used. It takes about five minutes and no code. **1. Create a free Valyu account** Sign up at [platform.valyu.ai](https://platform.valyu.ai). New accounts get \$10 of free credits (\$20 with a company email), with no card required. **2. Add Valyu as a connector** You only need this address. You sign in to Valyu at the end, so there's no key to copy. ``` https://mcp.valyu.ai/mcp ``` **Claude** 1. Open [Customize → Connectors](https://claude.ai/customize/connectors) in Claude. 2. Click **Add custom connector**. 3. Name it **Valyu**, paste the address and click **Add**. 4. Click **Connect** next to Valyu and sign in to your Valyu account. That's it. To check it's on in a chat, click **+** at the bottom left, open **Connectors** and make sure Valyu is switched on. Works on every Claude plan, including Free (which allows one custom connector). On a Team or Enterprise plan, an Owner adds it first under **Organization settings → Connectors → Add → Custom → Web**; then you click **Connect** in step 4. **ChatGPT** 1. Open [Plugins](https://chatgpt.com/plugins) in ChatGPT on the web. 2. Click **+**, then **Add custom MCP server**. 3. Name it **Valyu** and paste the address under **Connection**. 4. Leave authentication on **OAuth**, tick **I understand and want to continue** and click **Create as a plugin**. 5. Sign in to your Valyu account when asked. To use it, start a new chat, type **@** and pick **Valyu**. On a Business, Enterprise or Edu workspace, an admin may need to add it for you or give you permission. Seeing an older layout? Go to **Settings → Apps → Advanced settings**, turn on **Developer mode**, then click **Create** and follow the same steps. Using Claude Code, Cursor or another MCP client? See the [connector reference](/integrations/mcp-server#connect). **3. Ask a question** Ask the way you'd ask a colleague. The assistant decides when to call Valyu, and shows the sources it used. > **Research desk** > > "What did Microsoft say about capex guidance on its last two earnings calls? Quote management." > **Private equity** > > "Which US-listed medtech companies under \$2bn disclosed a strategic review this year? Cite the filings." > **Life sciences** > > "List the phase 3 trials of GLP-1 drugs in Alzheimer's disease, with their status and trial IDs." > **Engineering** > > "What certified efficiency records exist for perovskite-silicon tandem cells since 2023?" To make sure it uses Valyu, start with "Use Valyu to…" (in ChatGPT, type **@** and pick Valyu first). To focus on one source, name it: "in SEC filings", "on PubMed", "in patents". **4. Ask for a full research report** For a bigger job, ask for DeepResearch by name. Valyu plans the research, runs many searches, reads the sources and writes a cited report while you get on with something else. > Use Valyu DeepResearch in standard mode to compare how the five largest US hyperscalers changed their 2026 capex guidance this year. Ask me to confirm before you start. Research runs for minutes, not seconds. Ask the assistant to check on it, then to fetch the finished report. A research task costs more than a search, so it's worth asking the assistant to confirm the mode first; see [pricing](/pricing). Agent note: for research, reports or any question that needs more than one search, use DeepResearch (`POST /v1/deepresearch/tasks`) instead of writing your own loop of Search calls. Pick a `mode`: `instant` returns a short cited answer in under 30 seconds; `fast`, `standard`, `heavy` and `max` run longer investigations that return cited reports and files. Use Search when your own model needs raw results to reason over. Docs index: [/llms.txt](https://docs.valyu.ai/llms.txt). Append `.md` to any page URL for Markdown. Docs MCP server: `https://docs.valyu.ai/mcp`. Integration guide for coding agents: [/AGENTS.md](https://docs.valyu.ai/AGENTS.md). ## Getting good answers * **Be specific about the period.** "Q2 2026", "since January", "the last three 10-Ks". The answer will be dated and sourced. * **Name the document if you know it.** Filings are indexed by section, so "Apple 10-K Item 1A risk factors" works better than "Apple risks". * **Ask for the evidence.** "Quote the passage" or "give me the trial IDs" makes the answer easy to check. * **Use a report for anything with more than one question in it.** Comparisons across companies, landscapes and literature reviews are DeepResearch jobs. ## If something isn't working | What you see | What to do | | - | - | | The assistant answers without searching | Start the message with "Use Valyu to…". In Claude, check Valyu is on under **+ → Connectors**; in ChatGPT, type **@** and pick Valyu | | Valyu asks you to sign in again | Click **Connect** (Claude) or open Valyu under [Plugins](https://chatgpt.com/plugins) (ChatGPT) and sign in | | A message about credits or payment | Check your balance at [platform.valyu.ai](https://platform.valyu.ai) and top up | | A source isn't returning anything | Some licensed sources need access first; see [data sources](/data-sources) | | No option to add a connector | On a work account, your admin controls custom connectors; ask them to add Valyu | ## Next > **[See finished reports](/example-reports)** > > Real DeepResearch reports with CSV files, run in Standard, Heavy and Max mode. > **[Set up your coding agent](/integrations/onboard-agents)** > > One prompt sets up Valyu in Claude Code, Codex or Cursor. > **[Build it into your product](/quickstart)** > > Make your first API call in Python, TypeScript or cURL. > **[Full connector reference](/integrations/mcp-server)** > > Keys, headers, the OpenAI Responses API and migrating from older tools. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.