Skip to content

Your AI agent is confidently wrong about your revenue. Sigil fixes that.

Every team asks your AI agent the same question and gets a different number. Sigil makes them all get the right one. Built in production at a $13M Series A SaaS.
Get started free
No credit card. 2-min setup.
Claude
Hey Sigil, how is enterprise pipeline trending vs last quarter?
Used 2 integrations, loaded tools
Enterprise pipeline: $4.2M, up 18% vs Q2 at the same point (31 deals)
Enterprise defined as deals >= $100K ACV at companies with 500+ employees. Q2 same-day comparison: $3.56M across 26 deals.
The gap

Your agent reads every HubSpot field. It still gives your VP Sales the wrong number.

It doesn’t know that “qualified pipeline” means something different in your business, which is how your VP Sales ends up quoting $2.4M to the board when the real number is $1.87M.
Without Sigil
> "What's our qualified pipeline this quarter?"
$2.4M
# Counted test deals in "Test Pipeline"
# Included deals in "Negotiation" -- not a qualified stage here
# Used amount field instead of hs_acv
# Used createdate instead of hs_date_entered_qualified
# Double-counted deals associated with multiple contacts
21.9%
accuracy on revenue questions
With Sigil
> "What's our qualified pipeline this quarter?"
$1.87M
# Excluded test pipeline (pipeline_type != "test")
# Used hs_acv as deal value (not amount, not arr)
# Filtered to stages: "Discovery" through "Proposal Sent"
# Used hs_date_entered_[stage] for date filtering
# Deduplicated by deal ID, not contact association
99.5%
accuracy on revenue questions
Same CRM. Different answers. Your forecast is wrong and you don’t know it yet.
Proven in production

Built at a $13M Series A company, not a pitch deck

21.9%
99.5%
accuracy on revenue questions

Sigil started as an internal tool at a $13M Series A B2B SaaS with over a thousand HubSpot fields across three objects and 389 active workflows. Deployed company-wide as the primary self-serve analytics tool, it answers real revenue questions daily for every team.

The accuracy numbers come from testing real business queries against the data team's known-correct answers, before and after adding the context layer. Your board deck matches your CRM. First pull. Every time.

Measured across production business queries. Not synthetic benchmarks.

“60% of agentic analytics projects relying solely on MCP will fail by 2028 without a semantic layer.”

Gartner D&A Summit, 2026

“Most AI tools have access to data. What they don’t have is context.”

HubSpot CPO, Spring 2026 Spotlight
What it fixes

The four revenue problems your agent can't solve without context

Each one costs time, trust, or money.

Pipeline numbers that don’t match

Your VP Sales gets $2.4M. Finance gets $1.87M. Same CRM, different definitions. Sigil codifies what “qualified pipeline” means once, so every agent query uses the same stages, fields, and exclusions.

Board meetings with wrong numbers

The AI counted closed-lost deals in pipeline, used the wrong date field, and double-counted across pipelines. Nobody noticed until the board asked. One set of governed definitions prevents this.

New hire ramp time

60% of RevOps onboarding is hunting tribal knowledge: which pipeline is real, which 200 of 1,470 properties matter, what “qualified” means here. Sigil makes that context queryable from day one.

Lead scoring trained on noise

Your model trains on CRM data where “qualified” means different things to different reps and close dates are guesses. Sigil standardizes the definitions the model ingests. Signal in, signal out.

Get started free
No credit card. 2-min setup.
The context model

Four reasons your agent gives wrong revenue numbers

Each one eliminates a class of revenue errors. Together, they take accuracy from 21.9% to 99.5%.
{ metrics }

Metric definitions

What “win rate” or “ACV” actually means in your business. Which fields to use, how to calculate, what to exclude.
metric: win_rate
formula: closed_won / (closed_won + closed_lost)
exclude: partner_deals, renewals
field: hs_date_entered_closedwon (not close_date)
{ aliases }

Field aliases

Your team says "ACV" but HubSpot calls it hs_acv. Your VP Sales says "revenue" and means net ARR. Sigil maps the language gap so your agent uses the right property every time.
alias: “ACV”
maps_to: hs_acv
note: “not amount, not arr, not dealsize”
{ exclusions }

Exclusion rules

Test deals, internal accounts, churned-then-returned customers, the "Integration User" owner. The records that silently corrupt every pipeline report and forecast.
exclude_when: pipeline = “Test”
exclude_when: owner = “Integration User”
exclude_when: deal_type = “internal_test”
applies_to: all revenue queries
{ gotchas }

Known gotchas

The tribal knowledge that lives in someone’s head. "Always use fiscal quarters -- they start February." "Close date was unreliable before March 2024." "The partner pipeline has different stage names for the same stages."
warning: “close_date unreliable before 2024-03-01.
           Use hs_date_entered_closedwon instead.”
warning: “Partner pipeline stages map to Sales
           pipeline as: ‘Qualified’ = ‘Discovery’”
You describe these in plain English during a guided onboarding conversation. Sigil handles the structure. ~20 minutes, once.
How it works

Three steps to accurate answers

You wouldn’t hand a new RevOps hire HubSpot access and say “figure it out.” Give your agent the same onboarding.
01

Give it access

One line in your Claude config. Sigil runs alongside your existing HubSpot MCP. It doesn’t replace anything. It just adds meaning.

Your agent now has two tools: raw data access (your HubSpot MCP) and business context (Sigil). Takes 2 minutes.

02

Have the onboarding conversation

Sigil asks you the questions your new hire would ask:

“What counts as a qualified deal?”
“Which pipeline is the real one?”
“Do we include partner deals in win rate?”

Your answers become the agent’s permanent context. ~20 minutes, once.

03

Every question gets the right answer

When your VP Sales asks Claude about pipeline tomorrow, the answer matches the CFO’s board deck. That’s what a context layer does.

Every agent in your org now queries through your definitions. The right filters. The right properties. The right exclusions. No more “confidently wrong.”

How it fits

Sits alongside, never replaces

Your AI agent already has a HubSpot MCP for raw data. Sigil adds the context layer that makes those queries accurate.
Your HubSpot MCP

Raw CRM data

Deals, contacts, companies, properties. Everything in HubSpot, exposed to your agent as-is. 1,470+ properties. 200+ objects.

Sigil

Your definitions

Metric formulas, field aliases, exclusion rules, known gotchas. 47 business definitions that tell the agent what your metrics actually mean.

Your AI agent

Accurate answers

Combines raw data with your context. Every query uses the right filters, properties, and calculations.

Result
Your VP Sales and your CFO see the same pipeline number.
Your data stays in HubSpot. Sigil stores only your definitions: metrics, aliases, rules. It never sees your CRM data.
Common questions

Before you ask

Does Sigil access my HubSpot data?

No. Your data stays in HubSpot. Sigil only stores your business definitions: metric formulas, field aliases, exclusion rules, known gotchas. Your AI agent queries HubSpot directly through your existing MCP connection. Sigil never sees your CRM data.

How long does setup take?

2 minutes to add the MCP connection to Claude, ChatGPT, or Cursor. Then a guided onboarding conversation where you define your key metrics and business rules. Once.

What if our definitions change?

Update anytime. Every definition is versioned with full history. You can see what changed, when, and who changed it. Roll back if needed. Your agent always uses the current version.

Do I need engineering to set this up?

No. You add one URL to Claude, ChatGPT, or Cursor. Then you have a conversation in plain English about what your metrics mean. No code. No API keys. No schema files.

Does this work with Salesforce?

HubSpot has the deepest purpose-built onboarding playbook with pre-configured context for CRM objects, pipeline stages, deal properties, lifecycle stages, and custom properties. Salesforce is on the roadmap. Today, Sigil works with any CRM that has an MCP server via the universal onboarding flow. You'll define your own context structure instead of using the HubSpot-specific guided setup.

What about Gong, BigQuery, and other tools?

Sigil works with any tool that speaks MCP. Gong has a purpose-built playbook. 9 database engines (BigQuery, Snowflake, Postgres, and more) have dialect-specific playbooks. Every other tool uses the universal onboarding flow. See the integrations page for the full list.

From 21.9% to 99.5%. Twenty-two minutes of setup.

One connection. 47 definitions. Every agent query matches your data team's answers.

GET STARTED FREELOG IN
Free to start. No credit card. Works with 50 fields or 50,000.