Postfleet provides email infrastructure for AI agents, handling parsing, classification, and extraction of inbound emails to your schema, with built-in screening for prompt injection before agents process them.

How does Postfleet protect against prompt injection?

Postfleet screens all inbound emails for prompt injection attacks before they reach your AI agents, ensuring only safe and sanitized content is processed.

Can Postfleet extract data to a custom schema?

Yes, Postfleet can parse and extract information from emails to match your predefined schema, enabling structured data ingestion for your AI agents.

What types of emails can Postfleet handle?

Postfleet handles inbound emails of any type, automatically classifying and extracting relevant data based on your configuration.

Is Postfleet easy to integrate with existing AI agent workflows?

Yes, Postfleet is designed as a drop-in email infrastructure layer, seamlessly integrating with your AI agents to manage email processing and security.

Does Postfleet support real-time email processing?

Yes, Postfleet processes inbound emails in real-time, allowing your AI agents to receive and act on emails immediately after parsing and screening.

Postfleet - AI Email Assistant tools - Free trial, pricing intro, performance review, official site access and online experience

What is Postfleet?

Postfleet is an email infrastructure platform built specifically for AI agents. It ingests raw inbound emails, strips them of attacker-controlled content, and extracts structured data into typed JSON. The product acts as a trust boundary between the email wire and an agent's context, ensuring that only clean, classified, and injection-scanned messages reach the agent. Users deploy it to turn unpredictable email payloads into reliable, machine-readable inputs.

Application scenarios

AI agent inbox management
Route customer support or transactional emails directly into an agent's workflow without exposing it to raw HTML or hidden text.
Security-sensitive email parsing
Process invoices, receipts, or vendor communications where prompt injection or malware in attachments must be blocked before extraction.
Automated data extraction
Convert unstructured email content (e.g., newsletter, plain note) into structured JSON fields like `{vendor, amount}` with a confidence score.
Compliance and audit trails
Retain a full sanitization report, verdicts, and cleaned body for every message, useful for regulated industries.
Real-time email-to-API pipelines
Use the hosted MCP endpoint to create mailboxes, send/reply to emails, and wait for structured results in under 9 seconds.
Multi-client agent orchestration
Integrate with any MCP client (Claude, Cursor, Windsurf, Cline, VS Code) using the same API key and six tools.

Core Features

Seven-stage pipeline
From verification and deduplication through injection scanning and extraction, every email passes through deterministic stages before reaching the agent.
Email authentication parsing
SPF, DKIM, and DMARC results are surfaced alongside the message, giving you sender trust data without extra configuration.
HTML-to-text cleaning
Tracking pixels, styling noise, and quoted-reply chains are stripped deterministically, leaving only the meaningful body.
Sanitization and logging
Hidden text, zero-width characters, and homoglyph runs are removed, and every removal is logged in a readable sanitization report.
Virus and attachment scanning
Malware verdicts from SES gate delivery; attachments are scanned asynchronously at the byte level, and failed scans are never marked clean.
Spam classification and quarantine
A layered classifier (SES verdict plus a proprietary model) quarantines spam by default, with operator reclassification and redelivery.
Injection scan and extraction
A model reads the mail so your agent doesn't have to; injection-risk messages are retained without extraction, while clean mail becomes typed JSON with a confidence score.
Live demo mailbox
No signup or API key required—mint a real mailbox on the spot, send an email, and watch it arrive structured in the dashboard.
MCP integration
Works with any MCP client (Claude, Cursor, Windsurf, Cline, VS Code) via hosted API or local npx command, using the same six tools and key.

Target users

Postfleet is built for developers and teams building AI agents that need to process email safely. This includes engineers working on customer support bots, invoice automation, compliance workflows, or any system where raw email is a security risk. It also suits operators who need to inspect, reclassify, or audit email processing in a dashboard.

How to use Postfleet?

To start, open a demo mailbox directly on the Postfleet website—no signup or API key required. Send an email (newsletter, receipt, plain note) to that mailbox, and watch it arrive structured in the dashboard within seconds. For production use, integrate via MCP: use the hosted API endpoint (https://api.postfleet.ai/api/mcp) with a bearer token, or run the local MCP server via npx @postfleet/mcp. The same six tools (create_mailbox, send_email, reply_email, list_inbox, read_email, wait_for_email) work across all MCP clients.

Effect review

Postfleet delivers on its core promise: turning raw, attacker-controlled email into typed JSON with a confidence score, all in about 8.2 seconds. The live demo mailbox is a strong proof point—no mockups, no signup friction. The seven-stage pipeline (verification, authentication, cleaning, sanitization, virus scanning, spam gating, injection scanning) is thorough and fail-closed, meaning malware, spam, and injection-risk messages never reach the agent. For teams already using MCP clients like Claude or Cursor, the integration is immediate. The only missing detail is pricing, which isn't listed on the site, so you'll need to contact Postfleet for cost. Overall, it's a well-engineered trust boundary for any AI agent that needs to read email safely.

Postfleet