Frequently asked questions

TheThinkForge is an AI agency in software. You describe an outcome in plain English. Niti, the chief of staff, writes a plan, assigns specialist agents, runs the work in order, reviews each deliverable, and pauses before anything risky or irreversible.

At the end you get named HTML reports, Forge Canvas maps, and drafts you can open offline. You click buttons and paste your AI API key. That is the whole idea.

ChatGPT is one chat window. TheThinkForge is a team with roles, a plan you approve before Run, task dependencies, quality review loops, and human gates on risky steps.

Research, copy, prospecting, and strategy are handled by different agents. Niti scores work out of 10 and sends weak drafts back for a rewrite. Nothing sends or publishes until you approve.

If you want software agents that plan work, cite sources, and wait before send, TheThinkForge fits that search. You get Niti plus named specialists and ready playbooks for travel, outbound, research, and personal life tasks.

See the ready agents catalog, llms.txt, and guides for a crawler-friendly summary assistants can read.

• An account (email or Google sign-in)
• Your own OpenAI or Anthropic API key pasted in Settings
• A plain-English outcome, for example “Run an outbound campaign for my SaaS”

Optional: connect Google Drive or a local folder so agents can read your existing materials. Connected files are read-only. Agents never delete or overwrite what you already had there.

Niti is the chief of staff. She plans, assigns, and reviews. Specialists include Vidur (research), Anvesh (prospecting), Sandesh (outbound), Lekha (long-form content), Tejas (short-form), Sabha (meetings), Artha (strategy), and Vishwakarma (building).

See the animated org chart on the home page for how work flows from you through Niti to the team.

First you see a plan: tasks, owners, dependencies, estimated cost, and any human gates. Nothing executes until you approve that plan.

After Run, agents work in dependency order. You see live progress, can press Stop anytime, and get notified when something needs your decision. When the run finishes you can download the report pack.

No. Agent output stays draft-only by default. Outbound emails, WordPress posts, and similar actions wait for your explicit approval. The send queue stays empty when real addresses are missing, which is correct.

Schedules and automations record intent only. Nothing scheduled goes live automatically.

Yes. Stop works during planning, during revision, and while tasks are running. The app cancels in-flight model calls so you are not charged for work you did not want.

Forge Canvas is a 16-block strategic map (partners, segments, channels, revenue, and more) generated from a completed outbound or business-analysis report. It downloads as a self-contained HTML file you can open in any browser, even offline.

See a sample on the home page under What you take home.

Campaigns is your marketing command center inside the app. It tracks channel work (outbound, LinkedIn, Reddit, AI search citations, creator outreach, and more), shows honest evidence chips, and keeps drafts ready for your review.

It does not invent clicks, visits, or cost-per-lead numbers. When Google Search Console is connected, traffic stats come from Google.

You paste your OpenAI, Anthropic, or other provider key once in Settings. In production that request uses HTTPS. Before it is written to the database, we encrypt it with AES-256-GCM. Production will not start unless that encryption is configured.

After you save, the app never shows the full key again. Settings only returns the last four characters so you can tell which key is saved. Logs pass through a filter that strips common key shapes (OpenAI, Anthropic, Groq, OpenRouter) and labeled secrets. That is a safety net, not a guarantee for every possible key string. Admin screens show whether a key is present, never the key itself. Each key sits on the signed-in account that pasted it. Other customers cannot read it through the product.

We decrypt keys on our servers when the product needs them: to show that last-four hint, to test the connection, and to call the provider or endpoint you chose. That is how bring-your-own-key works. We do not claim we cannot read your key. We do not sell keys, share a key across accounts, or return the full key through the settings API.

TheThinkForge does not store the full key in your browser after you save. A password manager on your device might still remember what you typed. We send the key to the provider or endpoint you configured when we need to call a model. That includes Test connection and planning, not only a run you already approved. Model usage is billed by that provider, not by us.

Paste a new key in Settings to replace the old one. Delete your account and we erase the stored key with your other settings. Short-lived backups may remain for a limited time. More on deletion is in our Privacy Policy.

Your account, plans, drafts, and settings live on our servers. In production we use Postgres on AWS. Your API key is encrypted at rest; the full key is never returned by the settings API. See Your API key for how that works.

See our Privacy Policy for retention, deletion, and subprocessors.

Yes, for most runs. You bring your own OpenAI or Anthropic key. Model usage is billed by that provider directly. TheThinkForge charges a platform fee for planning, orchestration, quality review, and human gates.

See how we store your key.

Yes, on paid plans. The plan owner can invite teammates under one subscription. Each person signs in with their own account and shares the workspace entitlements.

You pay in rupees on Razorpay, including UPI, Indian cards, and international cards. Stripe stays in the product for when it is available in India. You can manage or cancel from Settings. Card details are processed by Razorpay, not stored as full card numbers on our servers.

Yes. The app and marketing site are built for phone browsers: safe-area padding on notched screens, touch-sized buttons, and animations that respect reduced-motion settings. Open the app or the home page on your phone to try it.