← Back to the journal

iHomefinder MCP: Where We Are, V1 to V5

A developer-to-developer status report on the iHomefinder MCP: what V1, V2 and V3 do, what we learned from the API, what is next, and how to connect during early access.

We are building an MCP server for iHomefinder accounts. This is a status report for developers and technical agents, not a launch announcement. It lists what is built, what we learned the hard way, what is still open, and where we would like your feedback.

Disclosure: IDX Webmasters is the IDX division of Virtual Results, which builds this MCP. iHomefinder is a trademark of its owner. This is an independent project, and nothing here implies that iHomefinder endorses or sponsors this MCP.

Access state: early access. The server is built and deployed, and it has its own connection URL (below). It is early software, so expect rough edges and tell us about them.

Versions at a glance

The server lives inside the same service as IDX Broker MCP, our IDX Broker MCP. A workspace connects to one provider, and a session only does real work for that provider.

  • V1 (!181, follow-up fix !182): leads, CRM writes, listings, hot sheets (read-only), reports. Built and deployed.
  • V2 (!183): digest, follow-ups, drafts, listing marketing, WordPress drafts, data quality. Built and deployed.
  • V3 (!184): export, open-house sign-in, buyer matches, tag manager. Built and deployed.
  • V3-infra (not started): scheduled digests, new-lead alerts, showing to calendar, CRM sync. Needs a scheduler and notifier.
  • V4 (planned): broker and team views, cross-provider lead merge.
  • V5+ (planned): needs vendor API changes or outside data.

The merge request numbers are from our private repository, so we list them as plain references. In total that is 36 iHomefinder tools for agents, plus 7 staff-only tools that Virtual Results uses to look after its own client sites.

What V1 does

V1 is the lead desk. It reads your account and, with confirmation, writes to it.

  • Lead inbox: all nine contact and request feeds merged into one timeline, with a "new since" filter.
  • Hot leads: leads ranked by score and recent activity.
  • Lead 360: one lead with profile, notes, tags, requests and report enrollments.
  • Writes, approval-gated: create a lead with tags and a note in one step (deduplicated by email), bulk create up to 25, update a lead, add or delete notes, add or remove tags, enroll a lead in a report.
  • Listings: featured listings, listings from a hot sheet, listing detail, and a listing kit for social posts that uses the listing URL on your own site.
  • Hot sheets: list them and read their listings. You can also build a hot-sheet-style search link on your own site.
  • Site activity: views, searches and saves, with period comparison.

Every write confirmation says whether the action will email the agent or the lead, because the vendor's behavior here differs by endpoint (see the quirks below).

What V2 and V3 add

V2 turns V1 data into daily work. Most of it is reading and composing, so it needed no new infrastructure.

  • Digest on demand: new leads, hot leads, unanswered requests, stale leads, and site stats against last week.
  • Follow-up queue and drafts: who has gone quiet, plus a draft email or text per lead based on their requests, listings and notes.
  • Listing marketing kit: captions, a description rewrite, open-house and price-change copy, with an attribution line and an IDX disclaimer placeholder.
  • WordPress drafts: a listing, neighborhood landing page or monthly update, shaped so it can be handed to a WordPress MCP as a draft (see below).
  • Quality and sources: duplicate and data-quality scan with merge advice, lead-source breakdown, and a "why are no listings showing" diagnostic.
  • Bulk note with a preview.

V3 adds:

  • Lead export as CSV or JSON, with header presets for Follow Up Boss and GoHighLevel imports.
  • Open-house sign-in: paste a sign-in sheet and get leads, tags and a note, with a cost estimate before it writes.
  • Buyer listing matches for one lead on demand, with a draft message.
  • Tag manager: usage counts, rename, merge and bulk tag.

What is working

  • All three versions are deployed to our production server, and each went through an independent review before merge. Blocking review findings were fixed before merge, and the rest in follow-up merge requests.
  • We tested live on Virtual Results demo accounts only. Notifications from those accounts go to a catch-all mailbox that we control. We did not run write tests on any customer's account.
  • Credentials and activation tokens never appear in tool output or logs, and text written by leads is marked as untrusted so an assistant does not treat it as instructions.

What we learned about the API

If you build against iHomefinder yourself, these may save you time.

  • Writes must be form-encoded. A note without a type is silently dropped. The tools always set the encoding and note type.
  • The lead list does not return score, source or address. Only the single-lead call does. Tools read single records for the top candidates.
  • Tags are case-insensitive. Adding a case variant is refused, and delete matches any case. The tag manager refuses capitalization-only renames so no tag is lost.
  • Creating a lead through the API sends no email. A contact request through the API emails the agent. Every confirmation says which one applies.
  • Enrolling a lead in a report may email the lead. Every confirmation notes it.
  • Hot sheets are read-only through the API. We read them and build search links instead.
  • The "OK for public display" flag can differ between the listing list and the single listing. The tools use the stricter value.

Why it has its own connection

OpenAI-based clients allow at most 128 tools per connection. Our shared connector was at 127, so iHomefinder's 36 tools could not be added to it. iHomefinder now has its own connection, and the iHomefinder tools no longer compete with IDX Broker's tools for that limit.

Roadmap

  • V3-infra: scheduled digests, new-lead alerts, showing requests to calendar events, and live CRM sync. These need a scheduler and notifier, and we have not designed them yet.
  • V4: broker and team views (we need a broker test account), and merging leads across providers.
  • V5 and later: features that need vendor API changes or outside data, such as a market analysis from MLS comps.

What we would love iHomefinder's API to add

These are open questions, not demands. Each one would remove a workaround.

  • Create or edit a hot sheet or saved search.
  • Create a lead without notifying the agent, for imports.
  • A consistent public-display flag between listing list and single listing.
  • A way to list and remove report enrollments.
  • Webhooks, or a "since" filter on lead and request feeds, so we can poll less.
  • Score and source on the lead list endpoint.

How to test it

Connect your own account. We never need your credentials by email, and we will not ask you to send them.

The connection URL is https://ihomefinder.virtualresultsplatform.com/mcp (36 tools, plus 11 WordPress tools). Every product follows the same pattern, https://<product>.virtualresultsplatform.com/mcp, so future tools get their own address the same way.

  1. Add it as a connector. In Claude, go to Settings, then Connectors, then Add custom connector, and paste the URL. In ChatGPT, add it as a developer-mode connector.
  2. Sign in with a Virtual Results account. Use the same login you would use for our other tools.
  3. Connect your own iHomefinder login using the settings link the tools give you. It is stored encrypted, and you never paste it into chat.
  4. Start with read-only prompts, such as "show me new leads since Monday", before you try a write.

If your iHomefinder account is managed by Virtual Results, use "Ask Virtual Results to connect it" on the settings page. If you want help getting set up, join the early-access list.

Connections: one per product

We now run one connection per product, such as IDX Broker MCP and iHomefinder, and each one includes the WordPress tools (36 plus 11, so 47 in total). Each stays under the 128-tools-per-connection limit that ChatGPT enforces, and all of them use the same Virtual Results login. We are also testing Cloudflare "code mode" to combine everything into a single connection. That is an experiment, and we are not promising it.

Using our WordPress MCP too

Virtual Results also runs a WordPress MCP for site content. It lists your sites, searches and reads posts and pages, uploads media, and creates drafts. Publishing needs a review step and two confirmations, so an assistant cannot make something live by itself.

It never uses WordPress passwords. A person signs in with their Virtual Results login, and the MCP acts as their own WordPress user through a short-lived signed pass that a small Virtual Results plugin on the site checks. Today that works on sites we host, which we link to your account. For other WordPress sites, a free "Virtual Results AI Connect" plugin is planned (coming soon). It would give you a one-click Connect and Virtual Results sign-in, with no passwords or API keys to paste.

What you need depends on who is asking:

  • A person: a normal user account, which is their Virtual Results login. It is not a service account.
  • An automation or bot: a dedicated service account, linked to a limited WordPress user.
  • A site Virtual Results hosts: we set up the access for you.

The WordPress tools are included in your iHomefinder connection, and the V2 drafts are built to hand to them, so a listing can go from your account to a WordPress draft you review from one connection.

What is most used, and what would make this easier

We do not publish usage numbers because we do not have enough testers yet. Our expectation, which we want you to confirm or correct, is that the lead inbox, hot leads and follow-up drafts matter most.

What we think would help most:

  • One-click connect, so you do not need to paste a connector URL.
  • Fewer tools per connection, so every client works. Each product is under the limit now, but a combined connection would not be.
  • Scheduled digests instead of asking each morning.

Feedback wanted

Tell us what is missing, what is wrong and what you would use daily. The early-access form is the best place, and feature requests go to the same team. If you are a developer who has hit the same API quirks, we especially want to compare notes.

Coming next for the platform

  • One Virtual Results account and dashboard across tools.
  • Billing built in. It is currently free during testing.
  • A share and referral program.

These are plans, not features you can use today.

The Virtual Results difference: What is a master reseller? →

Keep the conversation going.

Ask a question, share a workflow, or add your perspective. Comments are public and hosted on GitHub; a GitHub account is required to post.

Comment on GitHub