The field service platform that is open by design

# Find your workspace's regional API base
curl -X POST https://api.fieldproxy.ai/auth/org \
  -H "content-type: application/json" \
  -d '{"orgid": "<workspace-id>"}'
# => settings.url, for example https://api-india-1.fieldproxy.ai

curl -X POST https://api-india-1.fieldproxy.ai/api/read \
  -H "x-api-key: $FIELDPROXY_API_KEY" \
  -H "content-type: application/json" \
  -d '{"tableName": "your_table", "condition": {"id": 1}}'
# Reads the rows matching condition from one table

Endpoints, keys and automation behaviour: REST API reference and openapi.json.

1. Propose

   Every write, send and build is a dry run first: rows affected, before and after values, a pending_id.

2. Confirm

   Nothing is saved until a person agrees and confirm_write runs. Over 5 rows needs bulk permission; over 1,000 is refused.

3. Undo

   Every applied write returns an undo_id. Builds save as drafts, new automations and agents save switched off, and every call is logged.

// 1. The agent proposes; nothing is saved
propose_write({
  "summary": "Mark job 1042 as completed",
  "sql": "UPDATE jobs SET status = 'completed' WHERE id = 1042"
})
// => rows it would affect, sample before -> after values, and a pending_id

// 2. A person agrees, then the agent applies it
confirm_write({ "pending_id": <pending_id> })
// => the rows changed and an undo_id

// 3. Changed your mind?
undo_write({ "undo_id": <undo_id> })

How do I connect an AI agent to Fieldproxy?

Add Fieldproxy's MCP server to the agent. Your administrator switches on AI assistant access, you copy your workspace's MCP address (https://<api-host>/mcp, for example https://api-india-1.fieldproxy.ai/mcp in India or https://api-us-east-1.fieldproxy.ai/mcp in US East; it is shown in Settings, AI assistant when you create a key), and either sign in with OAuth from the client or create a key under Settings, AI assistant, New key and send it as Authorization: Bearer fpmcp.…. The agent then calls describe_schema to see your tables and starts working. It takes about two minutes.

Does Fieldproxy have a REST API?

Yes. Fieldproxy has a REST API for reading and writing workspace data: POST /api/read, /api/insert, /api/update and /api/delete on the workspace's regional API base (for example https://api-us-east-1.fieldproxy.ai), authenticated with an x-api-key header. Keys are created in Settings, Security, API keys, one per system, with an expiry, the operations they may use and an optional table allowlist. Writes through the REST API fire the workspace's automations. It is included on every plan.

Does Fieldproxy have an API for AI agents?

Yes. Fieldproxy is open by design. Besides the REST API, its MCP server (https://<api-host>/mcp, 170+ built-in tools) is an HTTP API any MCP client or agent framework can call, with tools for reading data, read-only SQL, proposed writes, messages, automations, agents and app building. Automations add inbound webhooks and outbound HTTP requests and signed webhooks, and public forms, portal users and CSV import and export cover the rest. API access is included on every plan.

Is Fieldproxy's data open to my own tools?

Yes. Each workspace runs on its own Postgres database. describe_schema shows every readable table and column, run_sql runs read-only SELECT queries on it, export_csv saves results as CSV, and automations can sync records to an external database with db_sync. Your records, screens and workflows are yours.

Which AI clients work with Fieldproxy?

Claude (custom connector), Claude Code, ChatGPT (developer-mode app), Microsoft 365 Copilot and Teams (through Copilot Studio), GitHub Copilot, Cursor and OpenAI Codex, plus any other client that speaks the Model Context Protocol over HTTP.

Can an AI agent change data in Fieldproxy safely?

Yes. Every write is proposed first with a dry-run preview and a pending_id, applied only by confirm_write after a person agrees, and returns an undo_id for undo_write. More than 5 rows needs bulk permission, more than 1,000 is refused, and keys can be limited by table, row, column and operation.

Can I send data into Fieldproxy from another system?

Yes. Call the REST API (POST /api/insert or /api/update with an x-api-key), or create an automation with a webhook trigger and POST JSON to its URL; the rule's steps create or update records from {{trigger.body}}. You can also import CSV or Excel files, and public forms take submissions from your website.

Can Fieldproxy call my systems when something happens?

Yes. Automation rules fire on record events, status changes, schedules, SLA timers, zones and more, and their http_request and send_webhook steps call any URL. send_webhook signs the body with HMAC-SHA256 in the X-Fieldproxy-Signature header so you can verify it.

Can I build my own screens and workflows through the API?

Yes. Fieldproxy apps are server-driven: every screen is JSON. Through the MCP build scope an administrator's AI can create apps, edit screens, add tables and columns, write automations and configure agents, as drafts or in a sandbox, then publish, with every version kept for rollback.

Does API or MCP access cost extra?

No. API access and the MCP server are included on every Fieldproxy plan. The only metered part is AI credits, which a few actions spend wherever they are started, for example running an insight or simulating an agent.