AI Assist
Freemium
Users prioritizing on-device processing and data privacy
Every inbox API on the market hands your agent raw, attacker-controlled email and calls it a feature — hidden text, zero-width characters, HTML your parser trusts and your model obeys, piped straight into context with full tool access. Postfleet puts a seven-stage pipeline between the wire and your agent so the payload that arrives is typed JSON, not an attack surface.
$20/mo · Free tier: 100 messages turned into JSON per month, 3 mailboxes, 5,000 sends per month, up to 100 sends per day · Checked Oct 23, 2026
The pipeline runs SPF/DKIM/DMARC authentication, strips tracking pixels and reply chains, removes hidden text and homoglyphs with a logged sanitization report, gates on malware and spam verdicts, then runs an injection scan before extraction ever touches your schema. Injection-risk mail is retained without extraction — it never reaches your agent. Clean mail becomes typed JSON with a confidence score in roughly eight seconds from external send to webhook delivery, per vendor-measured production data. Both hosted MCP and a local npx package are documented, and the same six tools work across Claude, Cursor, Windsurf, and VS Code. Operators can inspect raw versus processed output side by side in the dashboard for every message.
Bottom line: The right fit for a production agent that reads invoices or receipts from external senders — but if your pipeline requires processing injection-flagged mail rather than quarantining it, Postfleet's fail-closed design is a wall, not a dial.
Try it. No card.
Screened mail, small scale.
For a developer shipping agents.
For a team running agents in production.
View full pricing on postfleet.ai →
Pricing may have changed since last verified. Check the official site for current plans.
Find your fit
More options in Productivity. Compare pricing, platforms, and features side by side.
Freemium
Users prioritizing on-device processing and data privacy
Free
Users seeking culturally grounded AI responses
Auto-curated by the AIDiveForge recommendation matrix. These skills are predicted to enhance this tool based on category, capability, and domain signals.
Turn a transcript into decisions and assigned actions without inventing commitments.
Why: category partial · caps 0/0 · domain ops
Combine team standup notes while preserving blockers and ownership.
Why: category partial · caps 0/0 · domain ops
Turn a postmortem into an incident runbook with explicit verification gates.
Why: category partial · caps 0/0 · domain ops
Critique draft OKRs for measurable outcomes and controllable scope.
Why: category partial · caps 0/0 · domain ops
Postfleet is an inbound email infrastructure layer that sits between an external sender and your agent’s context. Each message moves through seven deterministic stages: sender verification and deduplication, SPF/DKIM/DMARC parsing, HTML-to-text cleaning, sanitization of hidden text and zero-width characters, malware scanning via SES verdicts, spam gating with a layered classifier, and a prompt-injection scan before schema extraction runs. The output delivered to your webhook is typed JSON — vendor, amount, confidence score — alongside a sanitization record and the full verdicts. The vendor describes the round-trip as 8.2 seconds from external send to structured payload, measured on production July 5, 2026.
The injection-screening stage is the architectural differentiator. A model reads the message before your agent does. If injection risk is detected, the message is retained and logged but extraction is skipped and the payload never reaches your agent. Hidden text removals, homoglyph runs, and every sanitization action are written to a report you can read — not silently dropped. That audit trail is what separates this from building a regex scrubber in front of a standard inbox API.
Postfleet fits teams building agents that accept email from untrusted external senders — invoice processors, receipt parsers, support intake pipelines — where attacker-controlled input is a real threat model, not a theoretical one. It does not fit use cases that require processing or acting on flagged mail without operator review; the fail-closed design is not configurable per the docs. Teams that need to route injection-flagged messages to a secondary agent rather than a quarantine queue will need to build that logic outside the pipeline.
Integration runs over MCP or REST. The hosted MCP endpoint requires only a bearer token; the local path runs via npx with no install beyond the API key in environment. The six documented MCP tools — create_mailbox, send_email, reply_email, list_inbox, read_email, wait_for_email — mean an agent can provision its own address and begin processing without dashboard access. A live demo inbox is available without signup, minting a real mailbox on request.
Sign in to rate this tool or suggest an edit.