ForumsQuestionsChatGPT Agent Access via API to Toodledo
ChatGPT Agent Access via API to Toodledo
| Author | Message |
|---|---|
|
gschlyter |
Kurt’s post on using ChatGPT to generate Toodledo CSV imports was extremely helpful, and it got me thinking about the next step. https://www.toodledo.com/forums/2/29104/-129180/read.html#jump
Has anyone here implemented direct AI/agent access to Toodledo using the API? Specifically, I’m interested in eliminating the CSV import/export workflow entirely and allowing an AI agent (ChatGPT, Claude, MCP server, etc.) to read/write tasks directly in Toodledo. I’ve found a couple of MCP/API projects on GitHub, but I’m curious whether anyone here is actively using this for GTD/task management. Would appreciate hearing about: Current setups/workflows MCP or API approaches Authentication challenges Mobile/voice workflows Reliability and limitations Feels like Toodledo could become an excellent backend for AI-assisted GTD if direct agent access is stable. |
|
kevinarth |
Also interested. I've been trying to get this to talk to Gemini but it seems Gemini isn't able to properly authenticate with the API to access my data dynamically. The workaround for now, which is a little clunky, is to have a daily digest email sent to me from Toodledo and then Gemini scrapes my email each morning to extract those items for inclusion in its Digest.
|
|
Steve |
I'm interested in enabling Codex to access my Toodledo info.
I asked Codex about this and here's what it said: >>> ...Yes, we could access it by setting one (MCP server) up. Toodledo has an official API v3 that supports OAuth2 and exposes tasks, notes, outlines, and lists. Their docs say tasks are available through the API, and auth requires registering an app, getting a client ID/secret, authorizing with scopes like basic tasks, then using access/refresh tokens. Sources: Toodledo API overview, Authentication, Tasks. Practical options: 1. Read-only quick connector: make a small local script/MCP server that lists tasks once you provide OAuth credentials. 2. Full MCP server: tools like toodledo_get_tasks, toodledo_add_task, toodledo_complete_task, maybe folder/context lookup. 3. Export/import fallback: if you can export Toodledo data as CSV/XML, I can analyze it locally without API setup. I can build the local MCP-style connector here, but I’d need either Toodledo API app credentials plus an OAuth flow, or an exported file. <<< Has anyone tried any of these? Gemini also had this to offer: >>> A Toodledo MCP (Model Context Protocol) connects your Toodledo task management to AI assistants like Claude or ChatGPT. It allows the AI to read your lists, manage contexts, and execute GTD (Getting Things Done) actions in real-time without needing you to write custom API code. You can set up and deploy a Toodledo MCP using the following platforms: - Zapier Toodledo MCP: Use Zapier's platform to bridge Toodledo's actions with any MCP-compatible AI tool without manual integration. - viaSocket Toodledo MCP: Automate tasks and allow AI assistants to make real-time updates to your Toodledo lists. - Community Servers: Open-source projects like the LobeHub Toodledo MCP Server let you manage tasks and goals through natural conversation. To use these servers, you will typically need a Toodledo API or App Token to authenticate your account. <<< This message was edited 1 day ago. |
|
Steve |
Here's an update: I asked Codex to try https://github.com/wwilson1017/toodledo-mcp
It told me that it found one caveat with that mcp: npm ci reported 12 vulnerabilities, including 1 critical, so before giving it broad write access to your task system, treat this as “usable but worth a quick audit.” Since then, Codex fixed the reported vulnerabilities, tested the code, successfully accessed my Toodledo info, and made two pull requests to the original project to share the fixes. Literally in minutes. Until those PRs are accepted in the main project, feel free to try my fixed branch here: https://github.com/lemketron/toodledo-mcp/tree/security/audit-and-oauth-hardening This message was edited 1 day ago. |
|
Steve |
There seems to be a slight bug in some Toodledo API responses (for create and others) -- the request works fine, but the response returns incorrect data. The request can be verified as being correctly completed by just fetching the task afterwards:
"One small quirk: the immediate create response showed incomplete/default-looking metadata, but a follow-up get_task returned the correct saved fields. So write access is good, and we found a minor “create response may not reflect final fields” behavior worth keeping in mind." |
You cannot reply yet
U Back to topic home
R Post a reply
To participate in these forums, you must be signed in.