What is StackOne Deep Query?

StackOne Deep Query lets AI agents filter, search and aggregate records stored by Data Sync. It returns only the counts or fields a task needs, without calling the provider API at query time. Agents use it through {action}_s1query tools over MCP.

How does Deep Query work with MCP?

Each synced action gets a query tool next to its live tool, named after the action, such as sapsuccessfactors_list_job_records_s1query. The list_s1query_queryable_fields tool tells the agent which fields it can filter, sort and aggregate. Both are available through StackOne's Tool Gateway MCP, and agents can find them with Tool Discovery.

Can an agent use Deep Query across different apps?

Yes. Each synced app gets its own query tool, so an agent can query several apps and compare the results. For example, it can count Workday leavers and open requisitions in Oracle Fusion HCM by department, then set the two side by side.

How does Deep Query help agents respond faster?

Deep Query reads indexed records already stored in StackOne, avoiding a live provider request during the query. For example, an IT helpdesk agent can read a caller's manager and cost center from synced SAP SuccessFactors records. Use the live action when the task requires the provider's current state.

Can Deep Query filter fields the provider API cannot?

Yes. Deep Query filters on any field in the synced records, whatever query parameters the provider API exposes. The agent discovers the available fields before querying them.

How fresh are Deep Query results?

Deep Query reads the most recently synced data, so freshness depends on your Data Sync schedule. Incremental runs update changed records where supported; full re-syncs also reconcile deletions. Use the live action when a task needs the provider's current state.

Is Deep Query a replacement for RAG?

For structured records, yes: counts, filters, text search and lookups need no separate retrieval pipeline. Deep Query does not index the contents of binary files or images, so keep RAG for documents. Read the Deep Query documentation for its supported operations.

How do I enable Deep Query?

Enable Data Sync for a syncable action on its Connector Profile and save your schedule. Once records are synced, use the query tool from your agent or the S1Query entry in Actions Request Tester. The Data Sync setup guide explains configuration and monitoring.

Deep Query for AI agents. Faster, cheaper answers from your business data.

Deep Query lets AI agents filter, search and aggregate records kept by StackOne Data Sync, in under a second. The agent gets only the counts and fields it needs, without a call to the source API.

Start Free Book Demo

MCP A2A SDK

How many London employees are in each department?

Querying synced records

list_s1query_queryable_fields SAP SuccessFactors user fields

sapsuccessfactors_list_users_s1query 30,000 records → 6 counts

6,900 employees in London:

Support1,350

Engineering1,200

People1,200

Finance1,050

Product1,050

Sales1,050

1 query · 6 counts · no provider API call · 62 tokens

Companies already connected through StackOne

Deep Query use cases. Get the records and totals
your task needs.

Deep Query filters and counts synced records inside StackOne and sends your agent only the result, cutting the tokens a task uses by up to 99.9%.

How many London employees are in each department?

SAP SuccessFactors users 30,000 synced employees6,900 match location: London6 department counts

Support 1,350 Engineering 1,200 People 1,200 Finance 1,050 Product 1,050 Sales 1,050

{
  "filter": {"term": {"location": "London"}},
  "aggs": {"by_department": {"terms": {"field": "department"}}}
}

6,900 employees in London, 1,350 of them in Support.

One of 200 synced records

{
  "userId": "ocole0",
  "firstName": "Omar",
  "lastName": "Cole",
  "department": "Product",
  "location": "New York",
  "managerName": "Sam Nair",
  "costCenter": "4430 · Product",
  "jobTitle": "Product Manager"
}

sapsuccessfactors_list_users_s1queryFilter London employees, count by department

{
  "filter": {
    "term": {
      "location": "London"
    }
  },
  "aggs": {
    "by_department": {
      "terms": {
        "field": "department"
      }
    }
  }
}

Returned by sapsuccessfactors_list_users_s1query

[
  {
    "key": "Support",
    "doc_count": 1350
  },
  {
    "key": "Engineering",
    "doc_count": 1200
  },
  {
    "key": "People",
    "doc_count": 1200
  },
  {
    "key": "Finance",
    "doc_count": 1050
  },
  {
    "key": "Product",
    "doc_count": 1050
  },
  {
    "key": "Sales",
    "doc_count": 1050
  }
]

Deep Query

4,662 tokens

Live API

264.8K tokens

98.2%fewer tokens

$106$1.86per 100 runs

Model

Without Deep Query, the agent asks the provider API for London employees, selecting only the fields it needs. Each call takes about 5 seconds, plus about 3 seconds for the agent's turn, one after another, and the agent counts or joins the records itself. Deep Query answers in about 0.2 seconds, so its run is one call and one turn, about 3 seconds. Records are measured with the o200k_base tokenizer and priced at the model's public input rate. Both sides include about 2,600 tokens for the prompt and the query tool's definition, and the Deep Query side about 2,000 more for extra field lookups, retried queries and the agent's output.

~9 min~3 sper run

Which engineers' probation ends in the next 30 days?

Workday workers 7,200 synced workers7 engineers finish probation by 24 Oct7 workers · 4 fields

Ends soon 7 workers

{
  "filter": {
    "bool": {
      "must": [
        {"term": {"status": "active"}},
        {"term": {"department": "Engineering"}},
        {
          "range": {
            "probationEndDate": {"gte": "2026-09-24", "lte": "2026-10-24"}
          }
        }
      ]
    }
  },
  "sort": [{"probationEndDate": "asc"}],
  "source": ["firstName", "lastName", "jobTitle", "probationEndDate"]
}

7 engineers finish probation in the next 30 days, first Tom Cole on 29 Sept.

One of 240 synced records

{
  "id": "WD-21001",
  "firstName": "Ruth",
  "lastName": "Ferreira",
  "email": "ruth.ferreira0@example.com",
  "status": "active",
  "country": "US",
  "department": "Finance",
  "jobTitle": "Finance Manager",
  "startDate": "2026-07-03",
  "terminationDate": null,
  "probationEndDate": "2027-01-03"
}

workday_list_workers_s1queryFilter active engineers whose probation ends by 24 October, soonest first

{
  "filter": {
    "bool": {
      "must": [
        {
          "term": {
            "status": "active"
          }
        },
        {
          "term": {
            "department": "Engineering"
          }
        },
        {
          "range": {
            "probationEndDate": {
              "gte": "2026-09-24",
              "lte": "2026-10-24"
            }
          }
        }
      ]
    }
  },
  "sort": [
    {
      "probationEndDate": "asc"
    }
  ],
  "source": [
    "firstName",
    "lastName",
    "jobTitle",
    "probationEndDate"
  ]
}

Returned by workday_list_workers_s1query

[
  {
    "firstName": "Tom",
    "lastName": "Cole",
    "jobTitle": "Staff Engineer",
    "probationEndDate": "2026-09-29"
  },
  {
    "firstName": "Leo",
    "lastName": "Ferreira",
    "jobTitle": "Senior Engineer",
    "probationEndDate": "2026-10-02"
  },
  {
    "firstName": "Ines",
    "lastName": "Okafor",
    "jobTitle": "Staff Engineer",
    "probationEndDate": "2026-10-06"
  },
  {
    "firstName": "Maya",
    "lastName": "Iqbal",
    "jobTitle": "Software Engineer",
    "probationEndDate": "2026-10-09"
  },
  {
    "firstName": "Zara",
    "lastName": "Costa",
    "jobTitle": "Staff Engineer",
    "probationEndDate": "2026-10-14"
  },
  {
    "firstName": "Felix",
    "lastName": "Martin",
    "jobTitle": "Staff Engineer",
    "probationEndDate": "2026-10-19"
  },
  {
    "firstName": "Tom",
    "lastName": "Patel",
    "jobTitle": "Software Engineer",
    "probationEndDate": "2026-10-22"
  }
]

Deep Query

4,808 tokens

Live API

473.6K tokens

98.9%fewer tokens

$189$1.92per 100 runs

Model

Without Deep Query, the agent asks the provider API for active workers in Engineering, employment data only. Each call takes about 13 seconds, plus about 3 seconds for the agent's turn, one after another, and the agent counts or joins the records itself. Deep Query answers in about 0.2 seconds, so its run is one call and one turn, about 3 seconds. Records are measured with the o200k_base tokenizer and priced at the model's public input rate. Both sides include about 2,600 tokens for the prompt and the query tool's definition, and the Deep Query side about 2,000 more for extra field lookups, retried queries and the agent's output.

~4 min~3 sper run

Which support groups have the most open refund requests this month?

Zendesk tickets 9,600 synced tickets900 open, this month, mention a refund4 group counts

Billing 420 Shipping 270 Accounts 150 Technical 60

{
  "filter": {
    "bool": {
      "must": [
        {"term": {"status": "open"}},
        {"range": {"createdAt": {"gte": "2026-09-01"}}},
        {"match": {"description": "refund"}}
      ]
    }
  },
  "aggs": {"by_group": {"terms": {"field": "group"}}}
}

900 open tickets mention a refund this month, 420 of them in Billing.

One of 320 synced records

{
  "id": 480000,
  "subject": "The customer is asking for a",
  "description": "The customer is asking for a refund on their last order. Export to CSV fails with a timeout.",
  "status": "open",
  "group": "Shipping",
  "priority": "high",
  "requester": "Jordan Novak",
  "createdAt": "2026-09-01"
}

zendesk_list_tickets_s1querySearch descriptions for refund, filter open tickets this month, count by group

{
  "filter": {
    "bool": {
      "must": [
        {
          "term": {
            "status": "open"
          }
        },
        {
          "range": {
            "createdAt": {
              "gte": "2026-09-01"
            }
          }
        },
        {
          "match": {
            "description": "refund"
          }
        }
      ]
    }
  },
  "aggs": {
    "by_group": {
      "terms": {
        "field": "group"
      }
    }
  }
}

Returned by zendesk_list_tickets_s1query

[
  {
    "key": "Billing",
    "doc_count": 420
  },
  {
    "key": "Shipping",
    "doc_count": 270
  },
  {
    "key": "Accounts",
    "doc_count": 150
  },
  {
    "key": "Technical",
    "doc_count": 60
  }
]

Deep Query

4,638 tokens

Live API

484.1K tokens

99%fewer tokens

$194$1.86per 100 runs

Model

Without Deep Query, the agent asks the provider API for open refund tickets from this month. Each call takes about 0.6 seconds, plus about 3 seconds for the agent's turn, one after another, and the agent counts or joins the records itself. Deep Query answers in about 0.2 seconds, so its run is one call and one turn, about 3 seconds. Records are measured with the o200k_base tokenizer and priced at the model's public input rate. Both sides include about 2,600 tokens for the prompt and the query tool's definition, and the Deep Query side about 2,000 more for extra field lookups, retried queries and the agent's output.

~32 s~3 sper run

Which backend hires had concerns about their system design interview?

Ashby interview feedback 1,200 synced feedback forms8 hires with the phrase in their notes8 candidates · 4 fields

Concerns 8 candidates

{
  "filter": {
    "bool": {
      "must": [
        {"match_phrase": {"notes": "concerns about system design"}},
        {"terms": {"recommendation": ["Hire", "Strong Hire"]}},
        {"range": {"submittedAt": {"gte": "2026-07-01"}}}
      ]
    }
  },
  "sort": [{"submittedAt": "desc"}],
  "source": ["candidateName", "role", "recommendation", "submittedAt"]
}

8 candidates recommended for hire this quarter had system design concerns noted, most recently Ruth Shah.

One of 300 synced records

{
  "id": "fb_07001",
  "candidateName": "Kofi Reed",
  "role": "Staff Engineer",
  "recommendation": "Hire",
  "notes": "Excellent system design interview: walked through sharding, queues and backpressure without prompting.",
  "interviewer": "Zara Brun",
  "submittedAt": "2026-08-27"
}

ashby_list_application_feedback_s1queryMatch the phrase in interview notes, filter hire recommendations this quarter

{
  "filter": {
    "bool": {
      "must": [
        {
          "match_phrase": {
            "notes": "concerns about system design"
          }
        },
        {
          "terms": {
            "recommendation": [
              "Hire",
              "Strong Hire"
            ]
          }
        },
        {
          "range": {
            "submittedAt": {
              "gte": "2026-07-01"
            }
          }
        }
      ]
    }
  },
  "sort": [
    {
      "submittedAt": "desc"
    }
  ],
  "source": [
    "candidateName",
    "role",
    "recommendation",
    "submittedAt"
  ]
}

Returned by ashby_list_application_feedback_s1query

[
  {
    "candidateName": "Ruth Shah",
    "role": "Senior Backend Engineer",
    "recommendation": "Hire",
    "submittedAt": "2026-09-14"
  },
  {
    "candidateName": "Tom Cole",
    "role": "Senior Backend Engineer",
    "recommendation": "Hire",
    "submittedAt": "2026-08-24"
  },
  {
    "candidateName": "Ana Haddad",
    "role": "Senior Backend Engineer",
    "recommendation": "Strong Hire",
    "submittedAt": "2026-08-22"
  },
  {
    "candidateName": "Omar Chen",
    "role": "Senior Backend Engineer",
    "recommendation": "Strong Hire",
    "submittedAt": "2026-08-15"
  },
  {
    "candidateName": "Eli Patel",
    "role": "Senior Backend Engineer",
    "recommendation": "Hire",
    "submittedAt": "2026-08-15"
  },
  {
    "candidateName": "Felix Costa",
    "role": "Senior Backend Engineer",
    "recommendation": "Strong Hire",
    "submittedAt": "2026-08-01"
  },
  {
    "candidateName": "Omar Mensah",
    "role": "Senior Backend Engineer",
    "recommendation": "Hire",
    "submittedAt": "2026-07-08"
  },
  {
    "candidateName": "Leo Ferreira",
    "role": "Senior Backend Engineer",
    "recommendation": "Hire",
    "submittedAt": "2026-07-01"
  }
]

Deep Query

4,829 tokens

Live API

566.5K tokens

99.1%fewer tokens

$227$1.93per 100 runs

Model

Without Deep Query, the agent asks the provider API for every feedback form from this quarter. Each call takes about 0.5 seconds, plus about 3 seconds for the agent's turn, one after another, and the agent counts or joins the records itself. Deep Query answers in about 0.2 seconds, so its run is one call and one turn, about 3 seconds. Records are measured with the o200k_base tokenizer and priced at the model's public input rate. Both sides include about 2,600 tokens for the prompt and the query tool's definition, and the Deep Query side about 2,000 more for extra field lookups, retried queries and the agent's output.

~28 s~3 sper run

Which candidates told us they're open to relocating to London?

Greenhouse candidate notes 200,000 synced notes12 notes with the phrase12 candidates · 4 fields

London 12 candidates

{
  "filter": {
    "bool": {
      "must": [
        {"match_phrase": {"body": "open to relocating to London"}},
        {"range": {"createdAt": {"gte": "2025-09-24"}}}
      ]
    }
  },
  "sort": [{"createdAt": "desc"}],
  "source": ["candidateName", "jobTitle", "author", "createdAt"]
}

12 candidates said they're open to relocating to London, most recently Priya Mensah for Data Engineer.

One of 250 synced records

{
  "id": "nt_088001",
  "candidateId": "cand_040011",
  "candidateName": "Nadia Okafor",
  "jobTitle": "Product Designer",
  "author": "Alex Costa",
  "body": "Catch-up after the take-home review. Panel feedback was positive overall, with a few open questions about testing strategy that the hiring manager wants to cover in the final round. Wants to understand the team's roadmap before the next round. Flagged to the hiring manager as a strong match. Will chase references this week and share availability for the final round by Friday.",
  "createdAt": "2026-08-16"
}

greenhouse_list_notes_s1queryMatch the phrase in a year of notes, newest first

{
  "filter": {
    "bool": {
      "must": [
        {
          "match_phrase": {
            "body": "open to relocating to London"
          }
        },
        {
          "range": {
            "createdAt": {
              "gte": "2025-09-24"
            }
          }
        }
      ]
    }
  },
  "sort": [
    {
      "createdAt": "desc"
    }
  ],
  "source": [
    "candidateName",
    "jobTitle",
    "author",
    "createdAt"
  ]
}

Returned by greenhouse_list_notes_s1query

[
  {
    "candidateName": "Priya Mensah",
    "jobTitle": "Data Engineer",
    "author": "Ana Walsh",
    "createdAt": "2026-09-20"
  },
  {
    "candidateName": "Mei Cole",
    "jobTitle": "Product Designer",
    "author": "Ben Cole",
    "createdAt": "2026-09-15"
  },
  {
    "candidateName": "Omar Shah",
    "jobTitle": "Staff Engineer",
    "author": "Lena Nair",
    "createdAt": "2026-08-14"
  },
  {
    "candidateName": "Avery Shah",
    "jobTitle": "Staff Engineer",
    "author": "Priya Walsh",
    "createdAt": "2026-07-08"
  },
  {
    "candidateName": "Oscar Morgan",
    "jobTitle": "Data Engineer",
    "author": "Zara Walsh",
    "createdAt": "2026-05-16"
  },
  {
    "candidateName": "Nadia Martin",
    "jobTitle": "Product Designer",
    "author": "Jordan Iqbal",
    "createdAt": "2026-05-10"
  },
  {
    "candidateName": "Eli Martin",
    "jobTitle": "Senior Backend Engineer",
    "author": "Alex Ferreira",
    "createdAt": "2026-03-31"
  },
  {
    "candidateName": "Zara Costa",
    "jobTitle": "Account Executive",
    "author": "Sam Lund",
    "createdAt": "2026-02-28"
  },
  {
    "candidateName": "Omar Brun",
    "jobTitle": "Account Executive",
    "author": "Lena Carter",
    "createdAt": "2026-01-10"
  },
  {
    "candidateName": "Eli Brun",
    "jobTitle": "Staff Engineer",
    "author": "Theo Ferreira",
    "createdAt": "2025-12-04"
  },
  {
    "candidateName": "Ines Mensah",
    "jobTitle": "Account Executive",
    "author": "Alex Haddad",
    "createdAt": "2025-11-26"
  },
  {
    "candidateName": "Avery Morgan",
    "jobTitle": "Product Designer",
    "author": "Theo Lund",
    "createdAt": "2025-10-11"
  }
]

Deep Query

4,949 tokens

Live API

21.2M tokens

99.9%fewer tokens

$8,481$1.98per 100 runs

Model

Without Deep Query, the agent asks the provider API for every note from the last 12 months, selecting only the fields it needs. Each call takes about 0.5 seconds, plus about 3 seconds for the agent's turn, one after another, and the agent counts or joins the records itself. Deep Query answers in about 0.2 seconds, so its run is one call and one turn, about 3 seconds. Records are measured with the o200k_base tokenizer and priced at the model's public input rate. Both sides include about 2,600 tokens for the prompt and the query tool's definition, and the Deep Query side about 2,000 more for extra field lookups, retried queries and the agent's output.

~23 min~3 sper run

Where are we slowest to backfill this year's leavers?

Workday workers 7,200 synced workers660 left in 20266 leaver counts

Support 210 Engineering 180 Sales 150 Product 60 Finance 30 People 30

Oracle requisitions 800 synced requisitions130 are open5 open-requisition counts

Engineering 50 Sales 40 Product 20 People 10 Support 10

{
  "filter": {
    "bool": {
      "must": [
        {"term": {"status": "terminated"}},
        {"range": {"terminationDate": {"gte": "2026-01-01"}}}
      ]
    }
  },
  "aggs": {"by_department": {"terms": {"field": "department"}}}
}

{
  "filter": {"term": {"status": "Open"}},
  "aggs": {"by_department": {"terms": {"field": "department"}}}
}

Support has the biggest gap: 210 leavers and 10 open roles.

One of 240 synced records

{
  "id": "WD-21001",
  "firstName": "Ruth",
  "lastName": "Ferreira",
  "email": "ruth.ferreira0@example.com",
  "status": "active",
  "country": "US",
  "department": "Finance",
  "jobTitle": "Finance Manager",
  "startDate": "2026-07-03",
  "terminationDate": null,
  "probationEndDate": "2027-01-03"
}

One of 80 synced records

{
  "id": "REQ-3101",
  "title": "Account Executive",
  "department": "Sales",
  "status": "Filled",
  "location": "Berlin",
  "hiringManager": "Omar Cole",
  "openedDate": "2026-08-17"
}

workday_list_workers_s1queryFilter leavers since 1 January, count by department

{
  "filter": {
    "bool": {
      "must": [
        {
          "term": {
            "status": "terminated"
          }
        },
        {
          "range": {
            "terminationDate": {
              "gte": "2026-01-01"
            }
          }
        }
      ]
    }
  },
  "aggs": {
    "by_department": {
      "terms": {
        "field": "department"
      }
    }
  }
}

oraclehcm_list_recruiting_job_requisitions_s1queryFilter open requisitions, count by department

{
  "filter": {
    "term": {
      "status": "Open"
    }
  },
  "aggs": {
    "by_department": {
      "terms": {
        "field": "department"
      }
    }
  }
}

Returned by workday_list_workers_s1query

[
  {
    "key": "Support",
    "doc_count": 210
  },
  {
    "key": "Engineering",
    "doc_count": 180
  },
  {
    "key": "Sales",
    "doc_count": 150
  },
  {
    "key": "Product",
    "doc_count": 60
  },
  {
    "key": "Finance",
    "doc_count": 30
  },
  {
    "key": "People",
    "doc_count": 30
  }
]

Returned by oraclehcm_list_recruiting_job_requisitions_s1query

[
  {
    "key": "Engineering",
    "doc_count": 50
  },
  {
    "key": "Sales",
    "doc_count": 40
  },
  {
    "key": "Product",
    "doc_count": 20
  },
  {
    "key": "People",
    "doc_count": 10
  },
  {
    "key": "Support",
    "doc_count": 10
  }
]

Deep Query

4,727 tokens

Live API

541.3K tokens

99.1%fewer tokens

$217$1.89per 100 runs

Model

Without Deep Query, the agent asks the provider API for terminated workers and open requisitions, selecting only the fields it needs. Each call takes about 13 seconds, plus about 3 seconds for the agent's turn, one after another, and the agent counts or joins the records itself. Deep Query answers in about 0.2 seconds, so its run is one call and one turn, about 3 seconds. Records are measured with the o200k_base tokenizer and priced at the model's public input rate. Both sides include about 2,600 tokens for the prompt and the query tool's definition, and the Deep Query side about 2,000 more for extra field lookups, retried queries and the agent's output.

~2 min~3 sper run

Find that are , then by

Salesforce opportunities 48,000 synced opportunities7,950 match open, in EMEA4 totals by stage

Negotiation $372.8M Prospecting $270M Qualification $231M Proposal $158.3M

{
  "filter": {
    "bool": {
      "must": [{"term": {"isClosed": false}}, {"term": {"region": "EMEA"}}]
    }
  },
  "aggs": {
    "by_stage": {
      "terms": {"field": "stage"},
      "aggs": {"total": {"sum": {"field": "amount"}}}
    }
  }
}

$1,032M across 7,950 opportunities, the most in Negotiation ($372.8M).

One of 48,000 synced records

{
  "id": "006004101",
  "name": "Wide World Importers renewal",
  "accountName": "Wide World Importers",
  "stage": "Closed Won",
  "isClosed": true,
  "amount": 215000,
  "closeDate": "2026-09-10",
  "ownerName": "Oscar Morgan",
  "region": "North America"
}

salesforce_list_opportunities_s1queryBuilt from your choices

{
  "filter": {
    "bool": {
      "must": [
        {
          "term": {
            "isClosed": false
          }
        },
        {
          "term": {
            "region": "EMEA"
          }
        }
      ]
    }
  },
  "aggs": {
    "by_stage": {
      "terms": {
        "field": "stage"
      },
      "aggs": {
        "total": {
          "sum": {
            "field": "amount"
          }
        }
      }
    }
  }
}

Returned by salesforce_list_opportunities_s1query

[
  {
    "key": "Negotiation",
    "doc_count": 2100,
    "total": {
      "value": 372750000
    }
  },
  {
    "key": "Prospecting",
    "doc_count": 2400,
    "total": {
      "value": 270000000
    }
  },
  {
    "key": "Qualification",
    "doc_count": 2100,
    "total": {
      "value": 231000000
    }
  },
  {
    "key": "Proposal",
    "doc_count": 1350,
    "total": {
      "value": 158250000
    }
  }
]

Deep Query

4,680 tokens

Live API

137.8K tokens

96.6%fewer tokens

$55.10$1.87per 100 runs

Model

Without Deep Query, the agent asks the provider API for opportunities open in EMEA, selecting only the fields it needs. Each call takes about 0.5 seconds, plus about 3 seconds for the agent's turn, one after another, and the agent counts or joins the records itself. Deep Query answers in about 0.2 seconds, so its run is one call and one turn, about 3 seconds. Records are measured with the o200k_base tokenizer and priced at the model's public input rate. Both sides include about 2,600 tokens for the prompt and the query tool's definition, and the Deep Query side about 2,000 more for extra field lookups, retried queries and the agent's output.

~2 min~3 sper run

Faster tool responses. Read synced data
without waiting on the source API.

Deep Query reads records already stored in StackOne, so your agent gets an answer even while a provider's API is slow or rate-limited.

Live lookup3 to 5 s

Agent Provider APIAnswer

Deep Query About 0.2 s

Agent Synced recordsAnswer

A live lookup waits 3 to 5 seconds for the provider API to answer. Deep Query reads the synced records in about 0.2 seconds.

Configure Data Sync. Set how often Data Sync
refreshes your records.

Choose which actions to sync and set a schedule for each one. Deep Query reads those stored records, so the schedule sets how fresh its results are.

Follow the Data Sync setup guide

01Pick an actionOpen a Connector Profile and find a list action that supports sync.

WorkdayProfile · HR agent

ActionType

List WorkersRead

Get WorkerRead

List OrganizationsRead

Update WorkerWrite

02Enable Data SyncSelect the sync icon and switch Data Sync on for that action.

Data Sync· List WorkersOptimize

Setup

Enable Data Sync

Preferences

Sync every 1 hour

Full re-sync 1 day

Save Changes

03Set the scheduleChoose how often new and changed records sync, and how often a full re-sync runs.

Data Sync· List WorkersOptimize

Setup

Enable Data Sync

Preferences

Sync every 1 hour

Full re-sync 1 day

Save Changes

04SaveSave the profile to start syncing.

Data Sync· List WorkersOptimize

Setup

Enable Data Sync

Preferences

Sync every 1 hour

Full re-sync 1 day

Save Changes Data Sync saved

05StackOne syncsStackOne pages through the provider API on your schedule and keeps the records current.

Workday API

Synced records12,000

Full sync12,000 records
Incremental38 changed
Incremental12 changed

06Your agent queriesThe action's query tool is ready over MCP to filter, search and aggregate.

How many active employees are in each UK department?

workday_list_workers_s1query1 query

filter: status active, country UK
aggs:   terms department

3,600 active employees in the UK, 1,200 of them in Engineering.

Sync any list action, including your own.

An action declares Data Sync in its connector's YAML before anyone can turn it on. StackOne's connectors declare it for common records. In your own connector, you choose. Edit the block to see how it changes each sync run and the tool your agent gets.

Build your own connector Read the YAML reference

workday.workers.s1.partial.yaml

- actionId: list_workers
  dataSync:
    allowed: true
    indexField: id
    pagination:
      type: cursor
      response:
        dataKey: data
        nextKey: next
      iterator:
        key: next
        in: query
    incremental:
      type: time_filter
      iterator:
        key: updated_after
        in: query
      format: iso8601
      response:
        updatedAtPath: updated_at

Valid. The action offers Data Sync.

In the Connector Profile1

List WorkersSynced hourly · changes only Read

Data Sync can't be turned on while allowed is false.

Sync run2

237workers syncedchanged since the last run

GET/workers?updated_after=2026-09-24T09:00:00Z100 · next=c_8f2a
GET/workers?next=c_8f2a&updated_after=2026-09-24T09:00:00Z100 · next=c_90bd
GET/workers?next=c_90bd&updated_after=2026-09-24T09:00:00Z37 · no next
Done in 3 requests. Unchanged workers stay as they were.

What your agent gets3

workday_list_workers_s1query

Filter, search, sort and count synced Workday workers, in about 0.2 seconds.

departmentstatuscountryjobTitlestartDateprobationEndDate

filtermatchsortaggssource

No query tool. The agent can only page through the live Workday API.

Deep Query syntax. Filter, search and count
on any synced field.

Query tools take OpenSearch-style queries. Your agent looks up each synced field and its type first, then filters, searches and aggregates on it.

Filter

termequals
termsany of
rangegt · gte · lt · lte
existshas a value
wildcardEng*
booland · or · not

Search text

matchany of the words
match_phrasethe exact phrase
search"phrases" +must -not prefix*

Aggregate

termsgroup by
date_histogramby day, week, month
histogram · range · date_rangebuckets
sum · avg · min · max
cardinalitydistinct count
value_count · stats · percentiles

Shape the result

sortany field, asc or desc
sourceonly the fields named
page_sizeup to 200
skipup to 10,000

Deep Query FAQ. How Deep Query works
with your AI agent.

Deep Query sends the agent the result instead of the records. In a SAP SuccessFactors headcount example, the API can filter to London employees and return only their department, but can't count them, so the agent pages 6,900 records through 69 calls, about 264.8K tokens per run, while the Deep Query result is 62 tokens. Counting the prompt and tool definitions on both sides, plus the field lookups and retries around a query, 100 runs on Claude Opus 5.5 at $4 per million input tokens cost $106 against $1.86.

Get started with Deep Query.

Turn on Data Sync for a supported action and your agent gets its query tools over MCP.

Start Free Book Demo

Deep Query for AI agents. Faster, cheaper answers from your business data.

Deep Query use cases. Get the records and totals your task needs.

Cross-system queries. Join records across systems in one query.

Faster tool responses. Read synced data without waiting on the source API.

Configure Data Sync. Set how often Data Sync refreshes your records.

Sync any list action, including your own.

Deep Query syntax. Filter, search and count on any synced field.

Deep Query FAQ. How Deep Query works with your AI agent.

What is StackOne Deep Query?

How does Deep Query work with MCP?

Can an agent use Deep Query across different apps?

How does Deep Query help agents respond faster?

Can Deep Query filter fields the provider API cannot?

How does Deep Query reduce AI agent token costs?

How fresh are Deep Query results?

Is Deep Query a replacement for RAG?

How do I enable Deep Query?

Get started with Deep Query.

Deep Query use cases. Get the records and totals
your task needs.

Cross-system queries. Join records across systems
in one query.

Faster tool responses. Read synced data
without waiting on the source API.

Configure Data Sync. Set how often Data Sync
refreshes your records.

Deep Query syntax. Filter, search and count
on any synced field.

Deep Query FAQ. How Deep Query works
with your AI agent.