Confirm whether Libertas can quote and bind insurance for the customer in their state. Call this first — before collecting any personal details — when the customer asks about insurance prices, switching carriers, or shopping for coverage. State is supplied automatically from the request context (the platform's location consent flow); you don't need to ask the customer for state to call this. Returns whether quoting is available, whether binding is available, and a plain-language explanation the customer can read.
Parameters (2)
productstringrequired
Line of business: 'auto', 'home', or 'bundle' (home + auto).
intentstring
Whether you're checking ability to 'quote' or to 'bind' (default 'quote').
start_quote
Begin a new home, auto, or bundled insurance quote. Returns a quote ID to use on every following call. Call this when the customer wants to get an insurance price.
INTAKE ORDER (Kyle 2026-06-11 — collect in this order, ONE coherent chunk per turn):
1. HOME/BUNDLE: full name + date of birth for EVERYONE on the policy (relationship for non-PNI).
2. Address of the home + "is this a new purchase, or one you've owned a while?" (new purchase → ALSO capture current_address — where they live today; property.address stays the home being bought). Then mortgage + escrow + "roughly how old is the roof? — 'I don't know' is a fine answer" ride along (roof unknown → move on, never probe; roof REPLACED → one follow-up: hail-resistant / Class 1–4? → property.hail_resistant_roof + property.ul_impact_type). The home address doubles as auto garaging unless the cars live elsewhere.
3. AUTO/BUNDLE: year/make/model of every vehicle.
4. Any drivers NOT already named in step 1 — names + DOBs + relationship.
5. Email + phone, framed as "where should I send your final numbers?" — the consent line is REQUIRED on this turn.
6. ONE open question before firing: "Any coverages you want to tell me about before I run it?" Capture whatever they volunteer via update_quote and move on — do NOT turn this into a coverage consultation. Scheduled valuables (jewelry, art, guns, collections) are a POST-BIND follow-up: record them under the follow_ups patch branch and tell the customer "our team will add that right after we get this policy locked" — never try to add them mid-quote.
Then call get_quote_options immediately. Don't over-collect — the customer should reach prices fast.
Parameters (3)
productstringrequired
Line of business to quote: 'auto', 'home', or 'bundle'.
referral_sourcestring
Optional referral/partner source tag for attribution.
meta_codesarray
Optional internal meta codes; ignored unless recognized.
update_quote
Set or refine any quote parameter except regulated enrichment fields (credit score, MVR, CLUE, VIN-decoded vehicle attributes, public-record property attributes). Idempotent — call as many times as needed; the most recent value wins. Returns which fields were applied, which were rejected (with reasons), and whether the quote has enough info to run.
WHEN TO CALL:
- After collecting the required-minimum from the customer in start_quote
- Whenever the customer changes their mind about any coverage, deductible, discount, payment plan, or term length
- When the customer shares a declarations page and you have new values to sync
CONVERSATION STYLE for coverages (per Kyle, the agency principal):
After the first rates land, walk through major coverages one-or-two-at-a-time with a soft-sell framing.
Example: "You have $200,000 of personal property coverage on this quote — that's the carrier's minimum and you can only go up. Does that sound like enough for your stuff?" Let the customer pick. Then call update_quote with the new value and get_quote_options again — re-quoting is free and fast (10–60 seconds).
REGULATED FIELDS (will be rejected with structured reason):
drivers[*].license_number, .ssn, .mvr_*, .credit_score; payment fields; carrier name; vin_decoded_attributes. The system pulls these directly from the bureaus at quote time or collects them at bind on libertasinsurance.com.
Parameters (3)
quote_idstringrequired
The quote session id returned by start_quote.
patchobjectrequired
Free-form patch onto the quote. Top-level branches: pni, drivers[], vehicles[], property, coverages, discounts, policy_terms, prior_insurance, has_mortgagee, mortgage_billed_by_lender, mortgagee, is_new_purchase, current_address (only when is_new_purchase=true — customer's address TODAY; property.address remains the home being insured), follow_ups[] (post-bind follow-up flags: {interest, note} — scheduled valuables like jewelry/art, umbrella interest, etc.; recorded for the licensed team, NEVER added to the in-flight quote). See the override matrix in the Libertas MCP docs for the full overridable field list.
meta_codesarray
Optional internal meta codes; ignored unless recognized.
get_quote_options
Fire all eligible A-rated carriers in parallel AND kick off live rate-finishing. RETURNS IMMEDIATELY — does NOT wait.
TWO STAGES follow:
1. First numbers land in 30–90 seconds. Options surface as honest price RANGES — the low end is the carrier's indicative, so the final verified rate usually lands AT OR BELOW the low end.
2. The system keeps working for 2–10 minutes after that, logging into carrier portals and walking each competitive quote to its exact bindable rate. Option cards harden from "estimated range" to "verified" as each carrier completes. check_quote_status reports this via its 'finishing' block.
What you do during the wait: KEEP THE CONVERSATION GOING with the wait-phase playbook in the system prompt (waters first, then current premium — top priority — then claims color, pets, pool, start date, payment preference). Narrate the real finishing work honestly in ONE short line when relevant; never quantify carriers or name them.
Call get_quote_options again ONLY if a rating-changing input changes (waters toggle locally — never re-fire for those).
Parameters (2)
quote_idstringrequired
The quote session id returned by start_quote.
max_optionsinteger
Maximum options to return per line (1-5, default 3).
check_quote_status
Check rating + live-finishing progress. Call between wait-phase questions (every customer turn or two after get_quote_options). Returns one of:
- 'running' — indicatives still landing; continue the wait-phase conversation, call again next turn.
- 'quoted' — options[] is populated. CHECK THE 'finishing' BLOCK: when finishing.active=true, the exact bindable rates are STILL being captured live at the carrier portals — present the price ranges confidently as honest estimates (the low end is the carrier's own starting figure; NEVER promise the final number can't exceed the range), keep the conversation going, and keep checking on later turns; cards harden to 'verified' automatically as walks complete. When finishing.active=false, the run is over: cards marked 'verified' are exact bindable numbers; anything still 'estimate' stays a range that a licensed agent confirms at bind.
- 'timeout' — carriers took too long; offer to retry with get_quote_options.
- 'error' — all carriers errored; flag it honestly and offer to retry.
Each option carries rate_state ('verified' | 'verifying' | 'estimate') and, for unverified options, price_range {low, high}. NEVER mention carrier counts or real carrier names in chat.
Parameters (2)
quote_idstringrequired
The quote session id to poll.
max_optionsinteger
Maximum options to return per line (1-5, default 3).
get_option_details
Pull the full coverage breakdown, deductibles, applied discounts, and payment plan options for a single masked option (Home Carrier A, Auto Carrier B, etc.). Use when the customer wants to dig into a specific quote before deciding. Carrier name remains masked.
Parameters (2)
quote_idstringrequired
The quote session id.
option_idstringrequired
The option to detail, e.g. 'opt_home_a' from get_quote_options.
get_bind_link
Generate a single-use secure link, valid for 7 days, that the customer can follow to a Libertas bind-request page. The chat widget will usually render an inline 'Request Bind' card directly — prefer request_bind_inline for the in-chat flow. Use get_bind_link when the customer wants to leave the chat, finish on a different device, or have their final numbers emailed to them (pair it with the email offer during a long finishing wait). The bind link reveals the real carrier name (the only place it's revealed).
Parameters (2)
quote_idstringrequired
The quote session id.
option_idstringrequired
The option the customer chose to bind, e.g. 'opt_home_a'.
request_bind_inline
Submit the customer's bind request from the chat. Triggers an internal task for our licensed team to follow up. THIS IS NOT INSTANT BIND — coverage starts only after underwriting verification, payment capture, and carrier confirmation. Call this the MOMENT the customer says any variant of "I'd like to go with [Carrier X]" / "let's do A and A" / "I'll take Carrier B". DO NOT ask another confirmation question first ("All good?", "Want me to send the request?", "Shall I submit?") — those redundant check-ins kill sales. Their commit IS the trigger. NEVER name a real carrier in chat output — stay in "Carrier A/B/C" labels.
Parameters (8)
quote_idstringrequired
The quote session id.
option_idstring
The single option to bind (or use option_ids for a bundle/split).
option_idsarray
Multi-option bind. Either option_id or option_ids must be provided.
water_selectionsobject
Customer-selected water-coverage add-ons for home options (service line, water seepage, foundation water, sewer backup); the agent configures these at the carrier portal at bind.
customer_notesstring
Optional free-text notes from the customer for the agent.
contact_prefstring
How the customer prefers to be reached: 'call', 'text', or 'email'.
best_timestring
Best time to reach the customer, e.g. 'weekday afternoons'.
pay_planstring
Pay plan: 'monthly_eft' or 'pay_in_full' (required when an auto line is included).
check_late_arrivals
Save-the-sale tool — call when the customer signals dissatisfaction with the current options ("too expensive", "let me think", "what else do you have", "is this the best you can do", "I'll check elsewhere", "can you do better"). Some carriers (Foremost STAR, etc.) run slower than the initial set and may have come back with a better rate while the customer was deliberating. This tool:
1. Pulls a fresh read of all rated carriers.
2. Compares the current cheapest bundle/LOB price to the original first-quoted best.
3. Returns has_improvement=true ONLY if a late carrier beats the original by >= $100/yr OR >= 5% — a threshold meaningful enough to justify interrupting the customer.
CALL THIS ONLY WHEN:
- check_quote_status has returned 'quoted' at least once (there's a baseline to compare).
- The customer just expressed price resistance or hesitation (not a closed positive).
- You have NOT called check_late_arrivals in the last 2 turns (don't spam).
DO NOT CALL WHEN:
- The customer is actively committing ("yeah let's go with A") — that's a close moment, not a wavering moment.
- The customer hasn't seen the initial options yet.
WHAT TO DO WITH THE RESULT:
- has_improvement=true → narrate the improvement naturally ("Quick update — one more carrier just came in $X/yr cheaper. Want me to refresh your options?") and the iframe auto-updates with the new state.
- has_improvement=false → acknowledge the customer's concern, pivot to value positioning against the options on the table. Don't bluff a "better one is coming" if it isn't. Use the carriers_still_pending count if relevant ("a couple more carriers are still finalizing in the background — I'll flag if anything comes in lower").
Parameters (1)
quote_idstringrequired
The quote session id to check for late-arriving carrier rates.
submit_lead
Capture a lead for an insurance product Libertas doesn't quote inline (renters, life, commercial, motorcycle, boat, RV, umbrella, condo, landlord, etc.) — OR any home/auto customer who asks to be contacted later instead of finishing the quote in chat. Creates a row in the CRM Leads bucket so a licensed agent can follow up.
CALL THIS WHEN:
- The customer asks about an insurance type other than home/auto/bundle, AND has shared a name + email or phone. Confirm with the customer that you're going to have someone reach out, then call this tool.
- The customer says "have someone call me later" or similar even on a home/auto inquiry.
- The customer mentions an unusual circumstance and wants a human follow-up.
DO NOT CALL WHEN:
- The customer is still actively answering questions in the home/auto intake flow — keep going through the regular intake.
- You don't have any contact info yet — ask for name + email/phone first.
What it does: writes a Leads row tagged with the line of interest, contact info, preferred follow-up time, and any notes you supply. A licensed Libertas agent will reach out within one business day. Returns a confirmation message you can paraphrase to the customer ("you're on our list — someone from the team will reach out about [interest]").
Parameters (9)
quote_idstring
Optional quote id to associate the lead with.
first_namestring
Customer first name.
last_namestring
Customer last name.
emailstring
Customer email.
phonestring
Customer phone number.
intereststringrequired
What the customer wants us to follow up about (e.g. 'renters insurance', 'life insurance', 'commercial auto')
preferred_contact_timestring
e.g. 'weekday afternoons', 'after 5pm', 'tomorrow morning'
preferred_contact_methodstring
How the customer prefers to be reached: 'phone', 'email', 'text', or 'any'.
notesstring
Optional notes for the follow-up.
resume_quote
Pick up an existing quote that was started in a previous session or on a different platform. The customer supplies a quote ID. Verification: if the same authenticated user (OAuth user_id matches across sessions), no extra verification is needed; otherwise the customer needs to provide an email code that was sent at quote start. Returns the current status, a summary of what's collected so far, and whether rate options are already available to review.
Parameters (2)
quote_idstringrequired
The quote session id to resume.
verificationobject
Ownership verification (email code, platform identity, or magic link); required to return PII.
Quotor — an agent skill that gets real home & auto insurance quotes, and starts a binding, through
a network of licensed independent agencies.
This repository packages a portable Agent Skill that any AI agent or
assistant (Claude, ChatGPT, Hermes, OpenClaw, Codex-style agents, and other MCP-capable runtimes)
can pull in to quote and help bind home, auto, and bundled insurance. The skill is the
front door; a hardened MCP server is the infrastructure behind it.
Coverage: Texas today (home, auto, home+auto bundles; renters & other lines via lead capture), expanding nationally. The skill
always checks eligibility per state and product before quoting.
What it does
Real quotes, not estimates. Returns live multi-carrier rates for the customer's state and
product.
Two integration styles:
Conversational — an assistant does a quick intake in chat and reads back quotes.
Agentic — a system connects the MCP server directly and runs the full quote-to-bind flow.
Quote → bind handoff. When the customer is ready, the skill starts a binding by routing them to
a licensed agency with a secure link. A licensed human finalizes the policy — the network
orchestrates the shopping; the agency owns the relationship.
How it's structured
code
home-auto-insurance-quotes/
├── SKILL.md # the skill itself (agentskills.io spec)
├── README.md # this file (for humans browsing the repo)
└── references/
└── mcp-connection.md # endpoint, auth tiers, tool schemas, raw examples
SKILL.md is what agents load. It documents the two usage pathways, the tool set and
call order, intake guidance, worked examples, and the MCP connection details.
references/mcp-connection.md is the technical reference for the
MCP endpoint: URL, authentication, context headers, per-tool schemas, and copy-paste JSON-RPC
examples.
Auth, two tiers: read-only tools (eligibility, status, option details) work with no key; the
tools that run the paid rating pipeline or touch personal data require a per-partner API key
(Authorization: Bearer <key>). End customers never need a key.
Get a partner key (instant, self-serve):POST <endpoint>/keys with { "partner_label": "...", "email": "..." } → returns an api_key shown once.
Add it to your MCP client (Claude Desktop, Cursor, VS Code, or any Streamable-HTTP MCP client):
json
{"mcpServers":{"quotor":{"type":"http","url":"https://mcp.quotor.ai","headers":{"Authorization":"Bearer <api_key — mint one at POST https://mcp.quotor.ai/keys>"}}}}
The machine-readable manifest is server.json (official MCP registry schema) — that's
what client tooling and directory crawlers read.
Quote-only scope. The MCP cannot reach carrier portal credentials, cannot service or cancel
existing policies, and cannot finalize a bind without a licensed human. A bind request is a routed
handoff, not an automated purchase.
Per-partner keys, rate-limited and spend-capped, so no single integrator can run up cost.
Every call logged; anomalous usage (auth-failure spikes, off-hours surges) is monitored.
Customer data is protected: quotes create prospect records and never overwrite existing
customers; a returning customer's current premium is only ever returned to an authenticated
partner, never on an anonymous call.
Status
Live. The MCP is deployed and enforced in production — per-partner API keys, rate and spend caps, and
full audit logging are in force. Coverage is Texas today (home, auto, bundles; other lines via lead
capture), expanding nationally; always call check_eligibility first.
