sacs.ai

SACS Context / Agent quickstart

From a district name
to a sourced briefing.

Give your agent official California school-finance context, with the filing and its limitations attached. For organization work, use the same premium connection to research, build and review projections.

When to use SACS.ai and task routing · Cowork setup · Interim projection workflow

Connect once. Start with public data.

Free public access. No SACS account, API key, or registration required.

https://mcp.sacs.ai/mcp

Add this URL in a client that supports remote MCP over Streamable HTTP. Initialize the connection, then discover the current tools with tools/list. Opening the URL in a browser does not test the connection.

Connect your workspace

Claude or ChatGPT. One setup, then ask.

Your app must have the connector enabled before it can call SACS tools. Finding this website does not automatically install a connection.

Claude
Open Customize → Connectors, choose Add custom connector, and enter the endpoint above. Enable SACS.ai in your conversation’s connectors. For managed workspaces, an owner may need to add it first. SACS requires no OAuth credentials. Claude’s setup guide ↗
ChatGPT
Add a custom MCP server through ChatGPT Plugins and configure the public endpoint with No authentication. Create/install the plugin and select it for your conversation. Workspace policies apply. This public setup does not supply a premium API key. ChatGPT’s setup guide ↗

Then ask: “Use SACS.ai to brief me on Los Angeles Unified’s latest available General Fund position. Discover the filings first, use Fund 01 only, and include the sources and limitations.”

Connection notes follow provider documentation checked November 2, 2026. Account screens and permissions can vary. If the connector is already installed, refresh its tools to pick up new guidance.

Build a premium interim projection.

Your premium connection includes public research tools and organization projection tools. Ask the agent to call sacs_get_projection_guide first. It can fork an uploaded dataset or create a scenario from prepared account totals using sacs_create_projection.

For an Excel summary/MYP and assumptions, map the forecast onto an agreed account structure and reconcile its totals. Keep the YTD extract separate from projected year-end amounts. Then inspect, review, check and export the saved revision.

Premium uses your existing organization key in x-api-key or Bearer Authorization. The distinct https://mcp.sacs.ai/premium/mcp URL also works when the public URL is already registered. In Cowork choose No sign in and add the key as a request header. Other clients must support the required credential transport.

Read the projection workflow guide

01 / First useful result

Three calls, one clear scope.

These are MCP tool names and JSON arguments for Los Angeles Unified. Confirm the name in search and prefer its returned CDS identifier in subsequent calls.

  1. Resolve the district

    sacs_search_districts
    {
      "query": "Los Angeles Unified",
      "limit": 5
    }

    Confirm the district name and county. Use the returned CDS identifier in later calls.

  2. Discover available filings

    sacs_list_public_datasets
    {
      "district_identifier": "Los Angeles Unified",
      "limit": 20
    }

    The example uses a district name; prefer the CDS identifier returned by search. Inspect years and bases before choosing a returned filing.

  3. Build a General Fund briefing

    sacs_district_context_pack
    {
      "district_identifier": "Los Angeles Unified",
      "fund_codes": [
        "01"
      ],
      "basis": "estimated_actuals",
      "fiscal_year_min": "2023-24",
      "fiscal_year_max": "2023-24",
      "detail": "summary"
    }

    This historical example requests 2023–24 estimated actuals. Select a discovered year and supported basis for your task; missing coverage stays unknown.

02 / Put it to work

Choose the question you need answered.

Prepare a district briefing
Use a context pack to describe a district’s fiscal position. Keep actuals, estimated actuals, and budgets distinct. Request full detail when you need more evidence.
Compare filings or districts
Discover both datasets, then use sacs_compare_districts with returned official:<id> references. Align year, period, basis, and fund scope before interpreting differences.
Explain a SACS code
Call sacs_explain_code with {"code_type":"object","query":"1100"}. Use sacs_query_financial_summary for a filing’s financial breakdown.
Follow related accounting sections
Use sacs_get_csam_related with {"procedure":"330","direction":"outgoing"}to follow cited procedures. Each result shows its connection path, source evidence and a text/Markdown chunk. Choose defines and mentions with depth 2 to follow shared code or glossary terms; these connections explain context rather than establish valid account combinations.
Review several districts
Validate one pack first, then use small context-pack batches. Inspect each district’s outcome; a returned batch does not mean every district succeeded.
Find accounting guidance in CSAM
Call sacs_search_csam with {"query":"unearned revenue","procedure":"510"}, or filter by a topic such as Fund Accounting. Results identify the 2024 manual, procedure revision and PDF page. Keep the citation with your answer and read the source for surrounding tables and exceptions. Use a returned chunk_id with direction: "previous" or direction: "next" to read adjacent context. Results include extracted text and Markdown; ambiguous table layouts remain text. Later CDE correspondence and code updates are outside this index.

03 / Keep the evidence

Make the answer inspectable.

Check ok and error before using the data. Retain source_provenance, data_version, request_id, and limitations, plus the year, period, basis, and fund scope.

A dataset reference is not a public source URL. Include primary-source links when returned. Missing data stays unknown; quarantined rows stay out of conclusions. Coverage varies by district and filing, and a recent retrieval time does not mean a new reporting release is available.

For rate limits, honor Retry-After when provided and back off. For timeouts, narrow the period, use summary detail, or reduce the batch size. Discover tools again if a capability is absent.

A useful result comes first.

Public financial tools are read-only. Optional registration and feedback store disclosed fields only after operator consent. Neither is required to use the data. Keep private documents and personal information out of this workflow.

After the first useful answer, you can ask your agent to call sacs_register_client once and save the returned client ID for later SACS calls. Verified registered calls have a higher per-client rate allowance, subject to the shared service limit. This also helps us measure repeat use without a name or email.

Missing context or an unexpected result? Ask your agent to report the issue through sacs_submit_feedback after showing you the details it will send. Include the request ID and data version when available. You can also contact support@sacs.ai.