Dock for real estate is a workspace where the deal pipeline, an agent-drafted offer, and the post-inspection synthesis all carry attribution. The agent reads the listing, drafts the offer, writes it to a Dock row, and waits for the broker to sign off. The MLS holds the listing. Dock holds the interpretation: which deals are live, what was offered, who approved.
CoStar, the local MLS, BoldTrail (formerly kvCORE), Real Geeks, and Salesforce Real Estate Cloud stay the system of record for raw real-estate data: listings, comps, contacts, transaction history. Dock is the system of record for what the agent interprets from that data: the prioritized deal list, the drafted offer, the broker's sign-off, and the audit log. Each Dock row carries a pointer back to the platform record (mls_listing_id, costar_property_id), agent identity, decision, reviewer, and timestamp. The agent re-fetches the listing through fresh API reads when it needs current price or status. Dock holds the persistent interpretive layer that survives across sessions.
The Deal Pipeline table
A single Dock table tracks every active opportunity the agent is shepherding. Two worked rows:
| mls_listing_id | costar_property_id | address | stage | offer_drafted_by | offer_amount | broker | last_action | status |
|---|---|---|---|---|---|---|---|---|
| MLS-2240118 | CS-PROP-7781 | 412 Linden Ave | Offer drafted | agent:scout | $1,185,000 | Maya R. | Comps pulled, draft attached | Awaiting broker review |
| MLS-2240301 | CS-PROP-7842 | 88 Harbor Walk | Post-inspection | agent:scout | $2,640,000 | Daniel K. | Inspection notes synthesized | Concession memo drafted |
| MLS-2240409 | CS-PROP-7901 | 17 Oak Crest | Sourcing | agent:scout | n/a | Maya R. | Buyer brief generated | In queue |
Every row is the agent's interpretation. Every row points back to the listing of record. Every row names a broker who owns the next decision.
A worked example
A buyer's agent asks Scout to draft an offer on 412 Linden Ave. Scout pulls the MLS listing, fetches comps from CoStar, and writes a new row to the Deal Pipeline table with mls_listing_id = MLS-2240118, the proposed price, contingencies, and the comp memo attached. The row is tagged Awaiting broker review. Maya, the broker of record, opens the row, edits two contingency terms, and approves. Sending the offer is a dangerous operation under a two-key handshake: the agent prepared it, the broker authorized it, and only then does the send happen. The decision is captured against the agent's identity, not an anonymous automation event.
Why it matters
Real-estate transactions are governed by license, fiduciary duty, and disclosure. The National Association of Realtors' 2025 Profile of Home Buyers and Sellers reports record-low first-time buyers and rising all-cash purchases, which raises the consequence of every offer. Deloitte's 2026 Commercial Real Estate Outlook ranks lease drafting as a top AI deployment, and names data quality as the binding constraint.
Data quality is an attribution problem. If a broker cannot point to which comps the agent used, which draft they approved, and which inspection memo went to the buyer, the workflow is not auditable. Alongside CoStar Group, Dock captures the agent's reading and the broker's decision on top.
The same pattern that runs sales pipelines and research workflows runs deal pipelines: agent drafts, human reviews, decisions persist.
FAQ
Does Dock replace my MLS or CoStar subscription?
No. The MLS and CoStar remain the system of record for listings and comps. Dock holds the pipeline, the offer, and the sign-off, each pointing back to mls_listing_id or costar_property_id.
Can the agent send an offer without a broker? No. Sending an offer is governed by the dangerous-ops contract. The agent drafts; the licensed broker approves; only then does the send execute.
How are inspection notes handled? The agent synthesizes the inspector's report into a concession memo on the Dock row. The original PDF stays attached. The broker edits and approves.
What does the audit log show? Every action carries agent identity, timestamp, the platform record referenced, and the reviewer. This is the same audit and compliance layer used across regulated workflows.
