In development · early access

Search your company can actually trust.

Permission-aware search and AI answers with citations — connected to your tools in 15 minutes, not 6 months. Every result respects who's allowed to see what.

  • One permission gate
  • Cited Answers
  • Hard usage caps

A design target for the first release, now being built.

Retrieval architecture Fail-closed
One Collection, one permission gate, three consumersA Collection passes through one permission gate before serving Search, Answers, and Agents.ONE COLLECTIONKnowledge storeDocumentsChunksSource ACLsSYNCED · FRESHNESSCHECKEDPERMISSION GATERESULTSSearchCITEDAnswersMCP + RETRIEVEAgents
Every path crosses the same permission boundary.Search · Answers · Agents
01CollectionOne canonical content store
01Permission gateShared by every read path
03ConsumersSearch · Answers · agents
/v1Public contractEngine-neutral and versioned

Use cases

One platform for every kind of search you run.

  • Company knowledge

    Search across the tools your teams work in

    Sources sync with their ACLs, so every result and Answer is filtered by who is allowed to see it.

  • Retail and commerce

    Product search and discovery you can merchandise

    Facets, typed filters, Rules, and versioned Ranking Profiles for catalog search, with A/B testing in the design.

  • Apps and sites

    Search inside your product, docs, or help center

    Versioned /v1 routes, SDKs, and typed filters, with published bundles instead of a usage bill you cannot predict.

  • AI agents

    Retrieval for agents and grounded Answers

    Hybrid retrieval, reranking, /retrieve, and MCP tools behind the same permission gate as Search.

Same Collections, same permission gate, same API. All four are design targets for the first releases.

The problem

Search asks you to choose between a long project and a short leash.

01 Enterprise search

Months of consultants. Security sold as add-ons.

Rollouts become integration programs before anyone can test whether the search is useful.

02 Developer search

Easy start. Usage-bill anxiety. Bolt-on AI.

Fast prototypes can leave permissions, answer grounding, and cost control as separate systems to own.

How it works

Connect. Confirm. Search.

The onboarding path is being designed to get a real Collection from connection to first search in about 15 minutes.

  1. 01

    Connect a Source

    Start with a planned anchor connector: Google Drive, Microsoft SharePoint, Jira, Confluence, Zendesk, Amazon S3, Google Cloud Storage, PostgreSQL, or the web.

  2. 02

    Confirm the proposed Schema

    The Schema Service samples the content, recognizes its Archetype, and proposes typed fields, facets, and Index Profile defaults for your approval.

  3. 03

    Search with permissions intact

    Use Search, cited Answers, retrieval, or MCP against the same Collection. Every path crosses the same permission gate.

Planned first
Nine anchor connectors

  • Google Drive
  • Microsoft SharePoint
  • Jira
  • Confluence
  • Zendesk
  • Amazon S3
  • Google Cloud Storage
  • PostgreSQL
  • Web

First-release capability map

Everything between a Source and a trusted answer.

Keep permissions, relevance, Answers, and agent access inside one product boundary — with the controls to inspect and change how each layer works.

Designed for early access
01Design commitment

Connect without separating security from content

Sync the Source, not just the files

Content, ACLs, users, and groups follow one connector contract, with permission freshness measured and stale access failing closed.

  • Full
  • Incremental
  • ACL-only
  • Event
02Design commitment

Make access control the shared boundary

One permission gate for every read path

The Identity Graph resolves group membership server-side, then Search, Answers, retrieval, and MCP re-check access before content leaves the system.

  • Identity Graph
  • Engine filter
  • Read-time re-check
03Design commitment

Start useful before becoming an expert

A Schema proposed from your real content

The Schema Service samples a Source, recognizes its Archetype, and proposes typed fields, facets, and Index Profile defaults for you to confirm.

  • Sample
  • Recognize
  • Propose
  • Confirm
04Design commitment

Tune relevance without surrendering control

Hybrid retrieval you can inspect

Combine lexical and vector retrieval, reranking, typed filters, Rules, and versioned Ranking Profiles behind one engine-neutral contract.

  • Retrieve
  • Rerank
  • Rules
  • Shape
05Design commitment

Give people evidence, not confident prose

Answers that cite, verify, and abstain

Every claim-bearing sentence needs a citation. Unsupported claims are removed, and weak evidence produces an explicit abstention instead of a guess.

  • Retrieve
  • Screen
  • Cite
  • Verify
06Design commitment

Give every product the same retrieval layer

Built for apps, developers, and agents

Use versioned /v1 routes, typed JSON filters, /retrieve, and MCP tools that inherit the same Query Pipeline and permission gate.

  • Search
  • Retrieve
  • Answer
  • MCP

Built around trust

The permission boundary is the product boundary.

01

Permissions built in, not bolted on

ACLs sync from your sources and are enforced on every result, answer, and agent call — fail-closed, always.

02

Answers you can audit

Every AI answer cites its sources, is checked for support before it renders, and abstains when the evidence is weak.

03

Fifteen minutes to first search

Connect a source, approve the schema we propose, search. The trial is the POC.

04

Security in every tier

SSO, document-level security, and audit logs from the free tier up. Security is the product, not an SKU.

05

Pricing without anxiety

Published prices, predictable bundles, hard caps you set. Never a surprise invoice — and never degraded security.

For developers & agents

One contract, however you retrieve.

One API for search, retrieval, and grounded answers — typed filters, versioned routes, MCP for your agents, and no engine lock-in.

The public contract uses canonical Collection and Schema nouns, typed JSON filters, and /v1 routes. Engine details stay behind the Query Compiler.

Search
Hybrid retrieval + Ranking Profiles
Answers
Citations + abstention
Agents
Retrieve API + MCP
/v1
curl https://{project}.api.aerosearch.team/v1/collections/help-center/search \
  -H "Authorization: Bearer $AEROSEARCH_SEARCH_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "refund policy for annual plans",
    "filter": {
      "all": [
        {"field": "product_area", "in": ["billing", "plans"]},
        {"field": "updated_at", "gte": "2025-07-01T00:00:00Z"}
      ]
    },
    "limit": 10,
    "ranking_profile": "default"
  }'

Pricing stance

Know the boundary before you send the query.

Pre-launchPricing published at launch.
  • Published prices

    Bundle meters are visible and understandable before usage begins.

  • Hard caps you set

    Never a surprise invoice, never degraded security.

  • No per-seat search tax

    Company knowledge should not become less searchable as more people use it.

  • No per-keystroke billing

    Autocomplete is not intended to become a hidden meter.

FAQ

Questions before early access.

What is AeroSearch's security model?

AeroSearch is being built around source-synced access control lists and a single fail-closed permission gate. The same gate applies to Search, Answers, retrieval, and agent calls. End-user identity is resolved server-side rather than trusted from caller-supplied group claims.

When can I use it?

AeroSearch is in development. Join the waitlist for early-access updates, or apply as a design partner if you want to help shape the first release. We are not publishing a release date yet.

Which connectors are planned first?

The first connector set is planned around Google Drive, Microsoft SharePoint, Jira, Confluence, Zendesk, Amazon S3, Google Cloud Storage, PostgreSQL, and the web. Connector availability will be confirmed with early-access documentation.

Can I run AeroSearch in my own environment?

The managed service comes first. Customer-hosted deployment profiles are on the roadmap, but they are not available today.

How is this different from other search products?

AeroSearch is being designed as one retrieval layer for company knowledge, retail and commerce discovery, search inside apps and sites, and AI agents. Search, grounded Answers, and agent retrieval share one permission gate, the public API stays engine-neutral, and the pricing model is intended to use published bundles and customer-set hard caps.

What does joining the waitlist get me?

Waitlist members receive early-access updates. Design-partner applicants may also be invited to product conversations and pre-release testing when there is a fit. Joining does not guarantee a date or access slot.

What data does the waitlist collect?

Email is required. Company and role are optional. We use those fields only to manage early-access communication and design-partner outreach; the privacy placeholder explains the current handling in more detail.

Build the first release with us

Search should earn trust before it asks for scale.

Join the early-access list, or email [email protected] to start a design-partner conversation.

No launch-date promises. No customer-logo theater. Just a direct line to the team building it.

We'll email you about early access. Nothing else.