MCP Server

Celeste Connector

Tools for querying firm data (DealCloud), browsing internal documents (SharePoint), and web research — all from your AI assistant.

DealCloud 7
SharePoint 3

Firm Data — DealCloud

Query, create, and update records in your DealCloud CRM

dc_list_objects

List Entity Types

Discover all available object types in DealCloud (e.g., Deal, Company, Contact).

Parameters
response_formatoptional"markdown" or "json" (default: "markdown")
Use this as the first step to understand what kinds of records exist before querying or creating data.
dc_list_fields

List Entity Fields

Return all field definitions for an entity type — names, types, required status, and validation rules.

Parameters
entity_typerequiredEntity API name, e.g. "Deal", "Company", "Contact"
include_choicesoptionalInclude picklist/dropdown values with their IDs (default: false)
required_onlyoptionalReturn only required fields (default: false)
response_formatoptional"markdown" or "json"
Always call dc_list_fields before using dc_find_records, dc_aggregate, or any mutation tools. Never guess field names — they vary per client.
dc_find_records

Find Records

Query DealCloud records with auto-detected mode: single record, text search, filtered query, or all records.

Parameters
entity_typerequiredEntity API name
entry_idoptionalRetrieve a single record by numeric ID
search_queryoptionalText to match against the Name field
filtersoptionalAdvanced query using operators: $eq $ne $gt $gte $lt $lte $in $nin $contains $startswith
field_idsoptionalSpecific field API names to return
limitoptionalMax records per page (default: 50, max: 1000)
offsetoptionalRecords to skip for pagination (default: 0)
order_by / order_directionoptionalSort by field API name; direction: "asc" or "desc"
created_after / created_beforeoptionalISO 8601 date filters
aggregateoptionalSet to "count" to return total count only
response_formatoptional"markdown" or "json"
You must call dc_list_fields for the target entity before using this tool. Using unverified field names in filters or order_by will result in errors.
dc_entry

Get Records by ID

Retrieve full entity records when IDs are already known, typically from a prior dc_lookup result.

Parameters
entry_idsrequiredList of numeric DealCloud entry IDs
Always pair with dc_lookup — look up first, then call dc_entry with the returned IDs for full record details.
dc_lookup

Natural Language Lookup

Search DealCloud using contextual phrases. Returns IDs and key fields for follow-up queries.

Parameters
keywordrequiredDescriptive search phrase, e.g. "Jupiter deal from Q3 2024"
Provide focused, specific information to identify and disambiguate records. After a successful lookup, always call dc_entry with the returned IDs — do not skip to dc_query.
dc_aggregate

Aggregate Records

Group and aggregate DealCloud records by a field value (count, sum, avg, min, max).

Parameters
entity_typerequiredEntity API name
group_byrequiredField API name to group by (verify with dc_list_fields first)
aggregate_functionoptional"count" (default), "sum", "avg", "min", "max"
aggregate_fieldoptionalRequired when using sum, avg, min, or max
filtersoptionalPre-filter records before aggregating
top_noptionalReturn only the top N groups by count (max: 100)
Field names and filter values vary by client — do not guess. Prefer aggregating without filters first to understand the data shape.
dc_query

Natural Language Query

Convert a plain-English question into a SQL query for data retrieval and analytics.

Parameters
queryrequiredSelf-contained natural language question, e.g. "deals with status active from December"
This is a stateless endpoint with no conversation memory. Each query must be fully self-contained — remove pronouns and references to prior context. Use only when dc_lookup + dc_entry are insufficient.

Document Management — SharePoint

Search, browse, and read files from connected SharePoint sites

search_files

Search Files

Search for files and folders in the connected SharePoint site by name or content.

Parameters
queryrequiredSearch term — filename, keyword, or phrase, e.g. "quarterly report", ".pdf", "budget 2025"
max_resultsoptionalMaximum results to return (default: 25, max: 100)
Use short, targeted queries of 2–4 key terms for best results. You can search by file extension (e.g. ".docx"). Results include item IDs for use with get_file_content or list_folder.
list_folder

List Folder Contents

Browse the SharePoint file structure by listing the contents of a directory.

Parameters
folder_idoptionalID of the folder to list. Omit to list the root document library. Folder IDs come from prior list_folder or search_files results.
Start with no folder_id to see top-level contents, then navigate deeper using folder IDs from the results.
get_file_content

Get File Content

Download a SharePoint file and extract its text content. Best for reading one file at a time.

Parameters
item_idrequiredThe drive item ID of the file (from search_files or list_folder results)
max_pagesoptionalFor PDFs, max pages to extract (default: 3, max: 20). Ignored for non-PDF files.
Supported formats: PDF (text extracted page by page), plain text, CSV, MD, JSON, XML, HTML, and most code files. Returns metadata only for unsupported formats (images, video, etc.).

Recommended Workflows

Common multi-step patterns for working with the Celeste connector

Look up a record in DealCloud

1
dc_lookup — search by name or description to get entry IDs
2
dc_entry — fetch full record details using the IDs

Query or filter DealCloud records

1
dc_list_fields — get verified field API names for the entity
2
dc_find_records — apply filters, sorting, and pagination using verified field names

Analyze DealCloud data

1
dc_list_objects — discover available entity types
2
dc_list_fields — get field names for the entity you want to group by
3
dc_aggregate — group and count/sum records by field value

Find and read a SharePoint document

1
search_files — search by document name or keyword
2
get_file_content — read the file content using the returned item ID

Browse SharePoint folder structure

1
list_folder — list root directory contents
2
list_folder with folder_id — navigate into subfolders using returned IDs
3
get_file_content — read specific files found during browsing

Example Prompts

Natural language examples mapped to the right tool for each task

ToolExample Prompt
dc_list_objects"What entity types are available in DealCloud?"
dc_list_fields"What fields does the Deal object have?"
dc_find_records"Find all active deals created after January 2024"
dc_entry"Get the full record for entry ID 12345"
dc_lookup"Look up the company Acme Corp in DealCloud"
dc_aggregate"How many deals do we have grouped by stage?"
dc_query"How many deals were closed in Q4 last year?"
search_files"Find the US Non-Hotel AM Update document"
list_folder"Show me what's in the Investments folder"
get_file_content"Read the IC Memo for Project Alpha"