Manus AI Email Research Agent

Email as a research source

Email threads contain decisions, pricing history, and agreements that no external search engine indexes. Manus AI turns an inbox into a searchable knowledge base by using the Nylas CLI to query messages, read full threads, and synthesize findings into structured answers — all without manual forwarding or copy-pasting.

According to a 2024 Radicati Group estimate, the average business professional receives over 120 emails per day. Most research tools look outward — web searches, document stores, databases — but some of the most valuable information sits in those inboxes: a technical spec attached to an email from six months ago, the exact wording of an agreement, or a pricing decision buried 40 messages deep in a thread.

Manus can cross-reference multiple threads, extract specific data points, and combine inbox findings with web research to produce a single briefing.

Prerequisites

Setting up Manus for email research requires three things: a Manus account, the Nylas CLI Skill configured in your workspace, and at least one connected mailbox. The full setup takes under 10 minutes if you already have a Nylas API key.

1. A Manus account. Sign up at manus.im. Free and paid plans are available.
2. Nylas CLI Skill installed. The Create a Manus Skill for Email and Calendar guide covers the SKILL.md file, setup script, and authentication for your Manus workspace.
3. At least one connected mailbox. You need a Nylas API key and a connected grant from dashboard-v3.nylas.com.

Search by keyword

Keyword search is the simplest way to find emails about a specific topic. Manus accepts a natural-language query, translates it into a Nylas CLI search command, and returns structured results — typically in under 3 seconds for mailboxes with fewer than 100,000 messages.

Prompt: "Find all emails about the Q4 contract"

Manus translates that prompt into a Nylas CLI search call. The --json flag returns machine-readable output so Manus can parse sender, date, and subject fields from each result. The --limit 20 flag caps results at 20 messages to keep response times fast.

nylas email search "Q4 contract" --limit 20 --json

Manus reads the results, identifies the most relevant messages, and presents a summary with sender, date, and subject for each match.

Read and summarize a thread

After finding relevant emails, Manus can read individual messages and produce summaries that highlight key points, action items, and deadlines. A single nylas email read call retrieves the full message body — headers, plain text, and HTML — giving Manus enough context to generate a concise summary in about 5 seconds.

Prompt: "Read the latest email from Sarah and summarize it"

Manus first searches for messages from Sarah using the from: prefix, then reads the most recent match by message ID. The two-step pattern — search, then read — keeps each API call lightweight and avoids downloading unnecessary message bodies.

nylas email search "from:sarah" --limit 5 --json
nylas email read MESSAGE_ID --json

The summary Manus produces includes key points, action items, deadlines, and any questions that need a response.

Extract decisions from a thread

Decision extraction turns scattered email threads into a chronological record of what was agreed, by whom, and when. Instead of reading through dozens of messages manually, Manus searches for relevant emails, reads each one, and compiles a structured decision log — a task that typically takes 15-30 seconds for a 30-message search set.

Prompt: "Find all emails about the product launch and list every decision that was made"

Manus first runs a broad keyword search with a --limit 30 cap, then reads each result individually using nylas email read. The --json flag on every call ensures Manus can parse sender names, timestamps, and message bodies programmatically rather than relying on plain-text pattern matching.

# Step 1: Search for relevant emails
nylas email search "product launch" --limit 30 --json

# Step 2: Read each relevant message
nylas email read MSG_ID_1 --json
nylas email read MSG_ID_2 --json
nylas email read MSG_ID_3 --json
# ... continues for each relevant result

The output is a research deliverable: a list of decisions organized chronologically, with references to the original emails including sender, date, and subject line.

Combine inbox and web research

Combined research merges internal email context with external web data in a single Manus session. Manus searches the inbox first, extracts specific data points like agreed prices or vendor names, then browses the web for comparable benchmarks — producing a briefing that draws on both sources. This two-phase approach typically yields 2-3x more actionable findings than either source alone.

Prompt: "Find what we agreed on pricing for Project X, then research current market rates"

Manus first searches your inbox for pricing discussions about Project X, extracts the agreed numbers, then browses the web for comparable market data. The result is a briefing that compares your internal pricing against current benchmarks.

Other examples of combined research:

• "Find the vendor proposal from Acme Corp in my inbox, then research their company background"
• "Pull all emails about the security audit and compare our findings with OWASP best practices"
• "Find what features our customers requested via email this quarter, then research which competitors offer those features"

SKILL.md for research workflows

A research-focused SKILL.md file tells Manus which CLI commands to use for search-and-summarize workflows. This 30-line file replaces the general-purpose skill with one optimized for inbox research — it omits calendar and contacts commands, which cuts Manus prompt overhead by roughly 40% and keeps the agent focused on email queries.

---
name: nylas-research
description: >
  Search and analyze email threads. Find messages by keyword, sender,
  or date range. Read and summarize threads. Extract decisions, action
  items, and key data points. Activate when the user asks about
  email research, inbox search, thread summary, or decision extraction.
compatibility: Requires internet access and a Nylas API key.
metadata:
  author: nylas
  version: "1.0"
---

# Nylas Email Research

Search and analyze email using the Nylas CLI.

## Setup

If `nylas` is not available, run:

```bash
curl -fsSL https://cli.nylas.com/install.sh | bash
nylas auth config
```

## Commands

### Search emails
```bash
nylas email search "QUERY" --limit 20 --json
```

### Read a message
```bash
nylas email read MESSAGE_ID --json
```

### List recent messages
```bash
nylas email list --limit 20 --json
```

## Research workflow

1. Search for relevant emails using keywords from the user's question.
2. Read the top results to extract details.
3. Compile findings into a structured summary.
4. If the user wants combined research, use web browsing after inbox search.

## Rules

- Always use `--json` for parseable output.
- Never send or modify emails during research tasks.
- Cite which email (sender, date, subject) each finding came from.

Tips for better email research

Search precision has the biggest impact on Manus research quality. Specific queries with sender names, date ranges, and multi-word phrases return 3-5x fewer irrelevant results than single-keyword searches, which reduces the number of messages Manus needs to read and speeds up the overall workflow.

• Use specific search terms. "Q4 contract renewal pricing" returns better results than "contract". Manus passes your keywords directly to the Nylas search API.
• Mention senders by name. "Find emails from Sarah about the budget" lets Manus filter by sender and keyword simultaneously.
• Set date ranges when possible. "Emails about the product launch from January 2026" helps Manus narrow the search and reduces credit usage.
• Ask for structured output. "List the decisions as a table with columns for date, decision, and who decided" gives Manus a clear format to follow.
• Combine with web browsing. Manus can research external context after searching your inbox. This is useful for competitive analysis, market research, and fact-checking.
• Batch related questions. Instead of asking five separate questions about a project, ask Manus to "compile a research brief on Project X covering timeline, budget, decisions, and open questions from my email."

FAQ

These are the most common questions about using Manus AI for email research with the Nylas CLI, covering security, search limits, and accuracy of AI-generated summaries.

Can Manus search my email without me giving it my password?

Yes. The Nylas CLI authenticates via OAuth or API key. Your email password is never shared with Manus. Credentials are stored locally in the Manus sandbox session and do not persist after the session ends.

How many emails can Manus search at once?

The --limit flag controls how many results are returned per search. You can set it up to 50 per call. For broader research, Manus can run multiple searches with different keywords and combine the results — in practice, 3-4 targeted searches with --limit 20 cover most research tasks without exceeding Manus session time limits.

How accurate are Manus email research summaries?

Manus works from the raw email text returned by the Nylas CLI, so quoted passages and factual details like dates, names, and numbers are exact. Summaries and decision extractions are AI-generated interpretations — always verify critical findings against the original thread before acting on them.

Next steps

• Email Analytics with Manus AI: analyze email volume, senders, and patterns
• Sync Email to CRM with Manus AI: extract contacts and push to your CRM
• Create a Manus Skill for Email and Calendar: full setup guide for the Nylas CLI Skill
• Command reference: every CLI flag, subcommand, and example
• Manus: Playbook reference — structure for chained-research playbooks Manus runs against email + web sources