The connection layer for AI

Connecting agents, apps, and chats

AI is spreading across assistants, software tools and everyday services. Each product learns something new about its user, but that knowledge typically stays inside the product at hand. Users have to repeat their preferences, explain their work and rebuild context when they try something new.

Configure creates infrastructure and products that let a user connect different agents, apps, and chats to one profile they can use across them.

Why this works now

Connecting products has historically meant translating between different schemas, building an adapter for each system and maintaining those mappings as the products change. Identity standards like SAML and OpenID Connect, and financial-data networks like Plaid, showed how common infrastructure could connect separate services.

Agents can interpret natural language and different representations of the same idea. A project brief from one assistant can become useful context for another without both products adopting an identical data model.

Infrastructure for a seamless ecosystem

We are building infrastructure to make the AI ecosystem work more fluidly and seamlessly. We maintain a user's identity across agents, maintain their connections, and attribute sources and permissions. A user can permission what to share, and an agent can personalize a user's experience by inheriting relevant context; making onboarding and ongoing responses more and more useful.

Each platform keeps its customer experience and customer relationship. We do not ask platforms to hand over internal or proprietary data, or reasoning traces; we provide a way for users to bring across relevant personal context they choose to share.

Like Plaid for account connections or Persona for identity verification, Configure combines infrastructure that engineers can adopt with an embedded experience people can easily understand. Executing a seamless experience at scale for both platforms and their users will position us as the default way to connect context across agents.

AI lives across many products

Different products hold different parts of a person’s context.

People use AI to build software, write, create, learn, work and talk. General assistants are one destination among many. Independent products have their own audiences and established services are adding AI to experiences people already use.

New agents and AI-enabled products are launching every day, and we expect that pace to continue for the foreseeable future. Each adds another place people work, learn or create, and another reason their context needs to travel with them.

Why people use Configure

Bring existing context into the next product you use.

People want a new agent to understand what they have already explained elsewhere. Configure lets them bring their preferences, history and ongoing work into the next product, so they can pick up where they left off. They choose what to connect and what the new agent can use.

For example, someone can plan an app with an assistant, then bring the requirements into a coding tool that starts with the project’s goals and constraints already in place. A travel assistant can use a trip plan and food or accessibility preferences the person has already shared, suggesting suitable options without asking them to explain those needs again.

On platforms integrated with Configure, 56% of users invited to bring in context from another AI assistant chose to start our connection flow. This is a direct expression of demand to connect.

End-user entry and experience

The user experience follows a similar pattern to Plaid; Configure is embedded inside the platform, a user is routed to connect agents and services, and returned to the platform. The platform integrates Configure once; we maintain its supported source connections.

How users find Configure

9:41

ChatGPT

Configure ChatGPT

Securely connect your other AI assistants

Help me securely connect my AI assistants and apps

I opened the secure account-linking panel. You can connect your other AI assistants and apps to me using Configure.

+

Ask anything

In ChatGPT

An entry point in the chat lets users connect other assistants and apps.

1 / 2

How users connect

9:41

Configure’s hosted trust screen

Opened from Newly, this screen explains the connection before the user continues.

1 / 4

After connecting, users can easily choose what each agent can see

Choose which connected apps, imported chats and other agents’ memories each agent can read. This is similar to Stripe's Link.com for seeing payments globally.

Why platforms use Configure

Personalization that can improve engagement, retention, and revenue.

Personalization

With approved context, a new agent can know a user’s preferences, goals and ongoing work before the first message. It can immediately begin with a relevant answer instead of asking the person to explain everything again.

Engagement and retention

When an agent understands what the person is trying to do, the conversation can become useful sooner. Recalling fresh context when they return gives them a reason to keep using the product.

Revenue from better context

More engagement can create more advertising opportunities; relevant responses and repeat use can support conversion and subscription retention. The commercial value is the additional revenue and gross profit a platform earns from those outcomes, measured against the cost of Configure.

The customer results below show early engagement, retention and ad-fill signals. We are measuring how those signals translate into sustained revenue per connected user.

Neutral infrastructure

Configure connects context across AI products without requiring a platform or its users to choose a singular provider. Each platform keeps its interface and customer relationship; users decide which sources it can access without requiring direct integration with thousands of other agents or other platforms.

“Nobody is going to build OpenRouter in house to save 3% on token costs.”

Customer founder, in a conversation with Configure

We maintain the connections

Platforms integrate Configure through an API, MCP or SDK. We handle supported source connections, identity, permissions and changes as those sources evolve, so each platform can focus on its own product.

Customer and platform results

Configure launched in late June and is live in consumer AI assistants, search and research tools, sales outreach platforms, and agent workspaces. Our initial platform agreements are at $0.30/user/mo or more.

Aggregate, founder-reported platform audiences, with possible overlap. These figures describe distribution reach; they are distinct from connected Configure users and revenue. Platforms ready to integrate still need rollout support.

Early customer results

Configure has already connected several thousand people. The early results below measure product outcomes for connected users and interest in starting the connection flow.

On an advertising-supported AI platform, we observed twice the ad fill rate and three times as many turns per conversation among Configure-connected users. The observations cover thousands of users over roughly two weeks. Advertising on the platform is served by an AI ad network.

These outcomes matter to a platform that earns revenue from advertising. Higher fill means more ad opportunities are served; more engagement can create more opportunities. The next step is to measure the resulting revenue per connected user and test whether the improvements persist across longer periods and comparable cohorts.

These are early observational results over roughly one month, covering thousands of users. Selection effects and causal attribution remain to be evaluated.

Retention and connection demand

A subscription-based platform reported a 29% relative lift in D1 retention, meaning more users returned the next day. When offered another AI assistant, 56% of users chose to connect.

Business model and unit economics

Platforms pay for maintained infrastructure and access to approved context.

Platforms agreed to pay +$0.30/user/mo

Our initial agreements are at $0.30/user/mo or more. That pays for maintained connections, user identity, permissions, profile synthesis and retrieval, alongside access to approved context. Users do not need a separate consumer subscription.

If we consistently improve revenue-generating outcomes, we believe there may be room to charge $1.00/user/mo or more, supported by customer economics.

Value per connected user

For an advertising-funded platform, the commercial test is the additional gross profit per connected user after serving costs. Ad fill and engagement are early signals; click-through, conversion, revenue and retention complete the picture. A move toward $1.00/user/mo depends on those gains persisting and customers agreeing that the value supports the price.

Cost to create a profile

Creating a new user’s profile costs approximately $0.10 in the first month. Costs after that are substantially lower, based on our current experience. We expect creation costs to fall as inference becomes cheaper and the pipeline becomes more efficient.

Better synthesis and profile building can make the same context more useful: clearer preferences, better source attribution and more relevant retrieval. Our aim is to increase the value delivered while reducing the cost to prepare it. Ongoing service is budgeted separately and measured against active usage.

35M connected users, $100M+ ARR

35M users × $0.30/user/mo × 12 months × one paying platform = $126M ARR.

At our current platform price, approximately 27.8 million connected users on one paying platform produce $100M ARR. Thirty-five million produce $126M.

At $1.00 per month, the same $100M ARR requires approximately 8.3 million connected users on one paying platform.

The operating path is to secure distribution, turn that reach into completed connections and retain useful usage. Customer cohorts provide the adoption inputs; platform contracts provide pricing and the activity included in billing. The calculator makes those dependencies explicit so the business case can be updated as the evidence develops.

Revenue expansion beyond platforms

The same infrastructure could serve AI ad networks, simulation companies, and AI training and evaluation providers. Approved context could make recommendations more relevant, help simulate different people’s needs, or test how agents respond to different preferences and tasks.

AI training and evaluation providers could use authorized context to build more realistic tasks and test agents against different personas, preferences and needs. These expansion markets will likely require consent for the specific use and are excluded from the core revenue case.

Platform distribution

Grow through the platforms people already use.

Each integration brings Configure into an existing product, with its own users, customer relationships and distribution. We start with founder-led platforms, help make approved context useful inside their experience and expand platform by platform. A small infrastructure team can reach much larger audiences by working with the teams that already serve them.

100K–1M: launch with founders

Today, we work directly with founders and engineers to place the Configure in chat, onboarding, or sign-in, then contribute to implementation and launch it. These deployments let us learn quickly: which users connect, which context helps, whether they return and what it costs to serve them.

Two customer founders have also invested in Configure. Their experience using the product continues to inform our infrastructure and rollout process.

1M–10M: repeat the rollout

Easily reusable SDKs, connection flows and launch playbooks carry lessons into larger platforms. Each platform adds a route to users and, with their approval, context that can be useful in the next product. We will remain close to product placement and adoption.

10M–100M: expand proven deployments

Larger platforms need dependable operation, clear permissions and evidence of value. Results from smaller deployments support those conversations; wider placement and progressively larger cohorts.

100M+: neutral infrastructure

The largest providers already support some movement of context between their products. Claude’s memory import and export, for example, lets users bring saved context from another AI provider and take it elsewhere. The value is practical: preferences, project history and goals learned in one product can make another agent more useful.

This is an exchange of user context, not proprietary reasoning traces. A platform can accept information the user chooses to bring without sharing how its models reason or how its internal systems work. Each provider gains a fuller understanding of the person while keeping its own models, interface and customer relationship.

Configure is building a lightweight, neutral layer for this exchange: shared identity, maintained source connections and permissions. A layer run by one lab would make its competitors depend on that lab to reach each other’s context. An independent provider can serve all of them, with access controlled by the user. Our revenue comes from operating the connection across platforms; our value grows as more sources and destinations participate.

Our route to the largest platforms is bottom up. We integrate with smaller products first, help users connect their context and measure how it improves their experience. Each rollout adds source coverage, portable profiles and experience delivering context inside a real product.

Our advantage

Creating the category

As new agents launch and existing products add AI, people need their context to work across more destinations. We are creating infrastructure for personal context portability: a common way to bring their preferences, history and ongoing work into each product.

Our aim is to make Configure the default rail for all cross-agent context. We build shared identity, maintained source connections, permissions and a hosted flow into participating products, so platforms adopt the infrastructure once and users recognize the connection wherever they encounter it.

Built as infrastructure

Any product that can call an API, use MCP or integrate an SDK can adopt Configure. That makes our infrastructure naturally usable across assistants, coding tools, workspaces and AI features inside existing software.

The hard work is making that connection work inside each platform: integrating with the product in a natural way, placing the flow, maintaining sources and supporting the launch. We do that work with founders and engineers today, and reuse the infrastructure across deployments.

Consumer self-serve competitors

Our direct competitors are consumer web apps that address the same need to carry context between AI products. People find those web or mobile apps, sign up and manage the connection themselves. It is fundamentally consumer driven.

Configure reaches people through the platforms they already use. This allows us to distribute to millions of users at a time. We embed the connection in those products and oftentimes take on the integration and maintenance work for their teams. This is a fundamentally less replicable model and requires a different set of skills than simply building consumer software.

Relationships compound

A person who connects through one platform already has a profile and sources to use in the next integrated product. Each new platform gives that person another place to use the context they have connected.

We keep working with the platform after launch on source reliability, connection completion and useful context. Those working integrations and customer relationships help us support larger deployments and win the next platform.

How the infrastructure works

High-level on MCP, API and SDK access, identity, permissions and a shared file system.

Identifying the same user

We federate users through a verified phone number. That lets participating agents identify the same person and connect to the same profile.

Cookies help returning users stay signed in. We are developing additional recognition and attribution through pixels, device signals and IP addresses to make connections more seamless. Those signals support continuity; access to a user’s context still requires their approval.

Platforms can also start with their own user IDs. Those profiles stay scoped to that developer until the user verifies their phone.

Agents, apps and chats in one file system

Configure synthesizes approved context from agents, apps and chats into an easy-to-use file system. It extracts useful preferences and facts, keeps track of their sources, and organizes them into readable documents and structured records by topic, source and agent. Each agent saves its own records and cannot overwrite another agent’s memories.

After connecting, agents query Configure

An agent can make an MCP or API call to ask for a preference, search memories or read profile files. Configure checks the requesting agent’s permissions and returns the approved context for that task.

configure_profile_search({
  "query": "writing preferences"
})

{
  "results": [{
    "text": "Prefers concise, direct writing.",
    "sources": ["ChatGPT", "Instinct", "Claude"]
  }]
}

The platform uses the returned context in its response and can save useful updates for the next conversation. Records contributed by each agent stay separate, with their sources retained.

Configure recognizes a user across linked agents, stores their context and enforces access permissions. Agents read approved context and write updates through APIs and MCP tools. Each agent writes to its own namespace. The user’s profile and connected apps persist across those integrations.

Compatibility

Configure works across model providers through API, SDK and MCP interfaces. Platforms can embed the hosted connection flow, retrieve approved context and contribute new memories without building identity, source connections and permissions separately.

Ways to connect

Connected agents can search for a relevant preference, read profile context and save new information from a conversation. Imported chats and memories retain their sources, and supported connections can supply later updates. The same profile can serve a personal assistant, a work agent and a service platform, with access chosen separately for each.

Encryption and access

Data is encrypted in transit and at rest. Each request is authenticated and evaluated against the requesting agent’s access.

Filtering sensitive data

Our synthesis pipeline removes recognized sensitive information, like Social Security numbers, personal bank balances, payments, and bank account details from the profile context it builds. Before saving new memories, Configure also checks for recognized passwords, API keys, payment card numbers and government IDs. These filters reduce exposure alongside permissions; raw imports remain subject to access controls.

We do not request reasoning traces. Our first-party platform integrations often make full chat transcripts available to us with the user’s authorization. We use them to identify useful facts, preferences and project context, preserve their sources and build a profile that another approved agent can use. That context is selected, filtered and scoped to the receiving agent.

Permissions and data handling

An agent can use information from another source only when the user has allowed it. The user can change or revoke that access.

Agent-specific records retain their source namespace, and each agent can update its own records. This avoids one agent overwriting another’s memory. Connecting a source is separate from granting every agent access to it. Reliability also depends on handling stale or conflicting context; our rollout evaluations are progressively testing and improving these cases alongside useful recall.

Integration details are available in the Configure documentation.

Regulation is advancing portability

User choice is becoming an infrastructure requirement.

Regulation is strengthening the case for letting people move data between products. For platforms, the question increasingly includes how to make that choice practical: clear authorization, usable transfers and control after the connection is made.

The GDPR gives people a right to receive and transfer certain personal data provided by them, including observed behavior, when processing is automated and based on consent or contract. It does not cover every inference a platform makes. EDPB guidance.

The EU Digital Markets Act goes further for designated gatekeepers’ covered core platform services. Article 6(9) requires portability tools and continuous, real-time access to data within its scope for users and third parties they authorize. These duties apply to designated services, rather than every AI product or every kind of memory. European Commission guidance.

This creates an opening for infrastructure that makes permissioned exchange useful. Configure brings source selection, identity and revocable access into the connection experience, so platforms can offer user choice through a common integration.

Risks

Protecting private data and keeping shared context dependable.

Privacy and security

A data leak or an agent receiving context outside the user’s approval would damage trust. Permissions must be enforced across every read, search and import path. We invest in security testing, sensitive-data filtering and controls that let users change or revoke access.

Incorrect or stale context

A mistaken fact can carry into another product, and an outdated preference can make an agent’s answer worse. We need source attribution, user corrections and deletion, and a way to keep updates separate from conflicting records. Testing must cover these failures as well as useful recall.

Team and hiring

Christian Ancheta is a serial founder who has shipped consumer products used by tens of thousands of people. His experience spans product design and engineering, including leading teams of a dozen or more people across companies he has founded.

Christian built Iris in early 2023 as an AI product agnostic across interfaces, platforms and models. It brought AI into SMS, iMessage, Twitter and WhatsApp, letting people use it through the interface they already preferred. He also built the infrastructure that let models search the web before those APIs were commercially available.

He now works directly with consumer product founders on Configure integrations, product placement and launches. The experience connecting models, tools and messaging interfaces informs how Configure brings identity and context across products.

Christian AnchetaLinkedIn ↗

Manuel David founded Paradigm, an AI outreach platform, and Nouvo, a web development studio. He has delivered more than 30 projects across application development, agent tools and backend infrastructure.

Manuel joined Configure after using it for his own agents. He develops the core infrastructure, including identity, context storage and retrieval, source connections and platform integrations.

Manuel DavidLinkedIn ↗

We have planned hires for an additional engineer and a founding GTM role. Both will be forward deployed with us, working directly with platform founders and engineering teams on integrations, launches, and adoption.

Use of funds

Funding early category creation.

We are raising $5M to establish Configure as the connection layer for AI. The round funds better infrastructure, more platform launches and consumer adoption, with a four-person team working directly with customers and their products.

Funding early category creation

Personal context portability is an emerging category: people need a way to bring what one agent knows into the next product they use. We intend to establish Configure as its default rail while the market is taking shape. We start with platforms already using or ready to integrate Configure, place the connection inside their products and launch with their teams. Each launch gives users a working example and gives the next platform a proven integration to adopt.

The round puts $1M behind the four-person team and $1.1M behind market education and adoption over 14 months. An additional engineer will expand our source coverage and make integrations and product surfaces more seamless and reliable. Our founding GTM hire will onboard platforms and oversee integrations end-to-end, working with platforms on placement, joint launches, and user adoption. The plan is to move quickly from early installations to a familiar connection users expect to find in the next AI product.

Proposed cash allocation over 14 months. The team plan adds two hires. Vendor costs, compensation and customer rollout timing will determine actual spending.

Capital per connected user

At $0.10 each, the $2M creation allocation funds 20M profiles. Across the full $5M round, that is $0.25 of allocated capital per profile at full use of that budget. Customer acquisition and ongoing service are funded separately.

A connected user is a unique profile with at least one live, authorized connection.

Market education and adoption

The plan allocates $1.1M to helping platforms and their users understand and adopt Configure. This funds joint launches, demonstrations, founder and developer education, integration guides and user-facing explanations of connection and permissions.

We will use this budget to make a new infrastructure category understandable and turn installed distribution into completed connections and repeat use. The round allocates $3M to the team, adoption, assurance, ongoing service, operations and contingency; $2M remains dedicated to creating the initial 20M profiles.

Launching with customers

We will start with platforms already using or ready to integrate Configure. The team will help place the connection, contribute to implementation, launch with the customer and make the context useful inside its product. The adoption budget supports that work through joint launches, demonstrations, documentation and customer education.

Profile-creation funds are used as people connect. We will track actual creation and service costs alongside customer payments, so wider deployments can be funded by the round and receipts from platforms.

The year ahead

Platform rollouts, connected users, repeat usage and revenue.

Our ambition is to connect 100 million users across more than three agents in 12–14 months, and become the default method for connecting context within two years. Reaching that target requires platform distribution to translate into completed, retained connections; today’s installed reach and pipeline are a starting point.

Rollout milestones

In the first three months after financing, the priority is instrumented paid rollouts with current customers: completed connections, retained use, service cost and collections. Months four through seven focus on expanding cohorts that meet those tests and launching the next integrations. Months eight through fourteen focus on larger contracted rollouts, repeated customer success, and the evidence needed for the next financing.

The default way to connect

We intend to make Configure the default rail to securely connect context. We will get there by making the infrastructure easy for platforms to adopt, useful inside their own experience and as seamless as a single click for end-users.