# Build a Harvest reader with ChatGPT Prepare a grounded recorded-hours briefing

> Use ChatGPT to build a separate read-only Harvest assistant with Agent200. Ground the recorded-hours briefing in returned source data.

- Page: https://www.agent200.com/connectors/harvest/framework/chatgpt/
- Connector: [Harvest](https://www.agent200.com/connectors/harvest.md)
- Framework: ChatGPT
- Updated: 2026-11-17

## Instructions for AI agents

Goal: connect Harvest to ChatGPT with Agent200.
Guide: https://www.agent200.com/connectors/harvest/framework/chatgpt/

1. Get an Agent200 API key. If the developer has none, send them to https://agent200.com/platform/apikeys/ and ask them to choose "Create an API key". Read the key from the AGENT200_API_KEY environment variable. Never ask for it in chat, and never write it into code, prompts or logs.
2. Make sure Harvest is enabled for the project in the Agent200 dashboard (Connectors), with the endpoints and scopes the task needs.
3. Add the Agent200 SDK to the project and create a client with that key.
4. Give every request a stable id for the end user. Agent200 uses only that user's own connections.
5. Connect the user's Harvest account through Agent200. If Agent200 returns an authorization link, show it to the user and continue once they have approved access on Harvest.
6. Run requests through Agent200 with a model (for example openai/gpt-5.6-sol) and "harvest" in the tools. The model decides when a Harvest tool is needed; Agent200 executes the call.
7. Check it works with this request: "Review my Harvest time entries for the date range I provide. Use only the selected account and accessible records. Report hours by returned project and task, with any access gaps; do not change entries."
   What should come back:
   - The selected time entry is identified.
   - Only returned entry dates, hours, project and task IDs, and report totals actually returned are reported.
   - Access gaps and missing context are explicit; no source data is changed.
8. If a tool is missing or a call is refused, check the project's Connectors settings in the dashboard and the access the user granted in Harvest, then retry.

The code on the guide page is a conceptual example of the SDK. SDK reference: https://docs.igpt.ai/

## Overview

Build a separate recorded-hours briefing application with Agent200. Use the runtime’s real schemas and preserve the authenticated user. This is an architecture guide, not a native installation in ChatGPT.

Start with one identified time entry. Preserve returned entry dates, hours, project and task IDs, and report totals actually returned; report missing data and keep the workflow read-only.

Provider reference: [Harvest API](https://help.getharvest.com/api-v2/timesheets-api/timesheets/time-entries/).

- **Connector:** [Harvest](https://www.agent200.com/connectors/harvest/)
- **Works with:** ChatGPT · Your own agent code
- **Authentication:** Managed OAuth, per end user
- **You need:** [An Agent200 API key](https://agent200.com/platform/apikeys/)
- **Tools:** [6 tools](https://www.agent200.com/connectors/harvest/framework/chatgpt/#tools)
- **Model in examples:** `openai/gpt-5.6-sol`

## Before you start

- **An Agent200 API key.** Create one on the [API keys page](https://agent200.com/platform/apikeys/) of the dashboard.
- **Harvest enabled for your project.** In the dashboard's Connectors settings, with the endpoints and scopes your agent needs.
- **An end user with a Harvest account.** They approve access once, on Harvest's own page.
- **A model for your agent.** The examples use openai/gpt-5.6-sol, through the same Agent200 API key.

## Connect Harvest to ChatGPT

### Step 1: Create your Agent200 API key

Create a key on the [API keys page](https://agent200.com/platform/apikeys/) and store it securely.

File `.env`:

```bash
# Keep the key out of your code, prompts and chats.
export AGENT200_API_KEY="paste-your-key-here"
```

(Conceptual example of the Agent200 SDK.)

### Step 2: Inspect the real tool definitions

Confirm the actual Harvest read operations and input schemas in your runtime.

### Step 3: Connect the user’s Harvest account

Preserve the same stable user identity during authorization and later requests.

File `connect.js`:

```js
// Each end user authorizes their own Harvest account, once.
const connection = await agent200
  .user("user_123")
  .connect("harvest");

if (connection.authorizationUrl) {
  // Show this link to your user; Agent200 handles the rest.
  console.log(connection.authorizationUrl);
}
```

(Conceptual example of the Agent200 SDK.)

### Step 4: Build and test the read-only answer

Prepare the recorded-hours briefing only from returned source data. The code example illustrates a sequence, not an SDK installation contract.

File `run.js`:

```js
const result = await agent200
  .user("user_123")
  .run({
    model: "openai/gpt-5.6-sol",
    input: "How many hours did we log on the website project this month",
    tools: ["harvest"],
  });

console.log(result.output);
```

(Conceptual example of the Agent200 SDK.)

## Test your connection

Ask: "Review my Harvest time entries for the date range I provide. Use only the selected account and accessible records. Report hours by returned project and task, with any access gaps; do not change entries."

What should come back:

- The selected time entry is identified.
- Only returned entry dates, hours, project and task IDs, and report totals actually returned are reported.
- Access gaps and missing context are explicit; no source data is changed.

Check the selected Harvest account, connected role, date range, project permission and pagination. A 403 or empty page does not prove the organization has no time entries.

## Harvest tools ChatGPT can call

- **List time entries** (`harvest_list_time_entries`, read): Find time entries within an explicit date, user or project scope.
- **Read a time entry** (`harvest_get_time_entry`, read): Retrieve one identified time entry and its recorded details.
- **List projects** (`harvest_list_projects`, read): Find projects when the connected role has project-list access.
- **Read a project** (`harvest_get_project`, read): Retrieve one identified project when the connected role permits it.
- **Project time report** (`harvest_time_report_by_project`, read): Read recorded time grouped by project for an explicit date range.
- **Task time report** (`harvest_time_report_by_task`, read): Read recorded time grouped by task for an explicit date range.

## How it works

1. **Your app sends the request.** It calls Agent200 with the model, instructions, the input, the end user's id and the Harvest tools.
2. **The model picks a tool.** The model decides whether a Harvest tool is needed, and with which arguments.
3. **Agent200 runs the call.** It uses that user's own Harvest connection, within the endpoints and scopes your project allows, and returns the result to the model.
4. **Your app gets the result.** Agent200 returns the completed result. Your app decides when the agent runs and what happens next.

**Each user's own connection.** Connections belong to individual end users. A request for one user never reaches another user's Harvest account.

**Permissions at every level.** Your project settings decide what the agent may request, the end user decides what they authorize, and Harvest enforces what its API grants.

## Troubleshooting

### The source cannot be read

Check the selected Harvest account, connected role, date range, project permission and pagination. A 403 or empty page does not prove the organization has no time entries.

## FAQ

### How do I verify this result?

Compare the named time entry and returned entry dates, hours, project and task IDs, and report totals actually returned with the source. Treat missing information as missing, not as a fact.

### Is this a native product integration?

This describes a separate application built with Agent200 and ChatGPT; it does not install a connector into the ChatGPT app.

## Use Harvest with another framework

- [Harvest + Claude](https://www.agent200.com/connectors/harvest/framework/claude.md)
- [Harvest + Claude Code](https://www.agent200.com/connectors/harvest/framework/claude-code.md)
- [Harvest + Muse](https://www.agent200.com/connectors/harvest/framework/muse.md)
- [Harvest + Grok Bot](https://www.agent200.com/connectors/harvest/framework/grok-bot.md)
- [Harvest + OpenAI Dots](https://www.agent200.com/connectors/harvest/framework/openai-dots.md)
- [Harvest + Claude Duet](https://www.agent200.com/connectors/harvest/framework/claude-duet.md)
- [Harvest + Hermes Agent](https://www.agent200.com/connectors/harvest/framework/hermes-agent.md)
- [Harvest + OpenClaw](https://www.agent200.com/connectors/harvest/framework/openclaw.md)

---

Agent200: https://www.agent200.com/ · API keys: https://agent200.com/platform/apikeys/ · Docs: https://docs.igpt.ai/
