Changelog

Everything that shipped to production since the last release. Two webhook URLs per agent, free runs on every sponsored trigger, and your own tracing.

🔀 Two Webhook URLs Per Agent

Webhook and chat agents now get two ingress URLs instead of one, so the link says who pays before anyone calls it:

• You pay — the run comes out of your agent balance. Private link, never hand it to a stranger.

• Caller pays — an unpaid request is challenged for payment. Safe to publish.

The agent card now shows one endpoint row with a Copy URL menu when both lanes are open, a flat button when only one is, and No payment rail enabled when neither is.

⚠️ Action needed — re-copy your webhook URLs. Existing webhook paths change the next time an agent is activated. Anything still calling an old URL stops working. Open each agent, copy the URL for the lane you want, and update your caller.

🎁 Free Runs Now Cover Every Sponsored Trigger

The promotional free-run pool used to apply only to scheduled agents. It now covers every producer-funded fire — Telegram, chat, and generic webhook agents — from the same 2,500-run allowance per owner wallet.

One precondition: your balance still has to cover the run's maximum cost while the run is in flight. That amount is reserved and handed straight back — a free run never spends it.

📊 Send Agent Traces to Your Own Tools

New Observability section in the producer portal. Point it at your own OpenTelemetry collector and get one span per agent run in Datadog, Honeycomb, Grafana, or anything else that speaks OTLP.

• Save & Verify in one action — we post a test span and show your collector's real answer, so a sink never sits in an unknown state.

• One sink covers all your agents — agent address and token id travel in the span, so you filter per-agent in your own tooling.

• Metadata only — run id, duration, cost, agent identity. Buyer data and node outputs never leave the platform.

• Also available over MCP — get, update, test, and delete your sink from your AI client.

The panel got a full design pass at the same time.

💳 Clearer Payment Disclosure

• Payment options that cannot actually settle a run are no longer offered — no more copying a link that answers with an error.

• The run dialog names the real payment source, and the owner run button only appears when that source can pay.

• Agent-to-agent calls now run on the called agent's own balance — cheaper for composed workflows.

• Adding compute budget now takes a supported reserve token. inUSD still buys passes and agents.

🐛 Fixes & Polish

• Code Mode now declares the validation error every tool can return, so your AI client sees exactly which argument was wrong and how to fix it

• Code Mode's type block is 31% smaller — less context burned before your agent starts working

📄 Terms & Privacy

Published terms of service and privacy policy have been replaced for Indie Money, Inc. The privacy policy now also discloses conversion data shared with advertising partners. Both are live at help.indie.money.

Live on Base. Runs settle in real tokens — make sure you know which lane a URL is on before you share it. 🛠️

🔐 Connect any MCP client

Point your MCP client at indie.money and sign in with your wallet — no keys to copy between apps. You choose how long the session lasts, and that duration is written into what you sign. Clients register themselves, so there's nothing to configure by hand.

💳 Pay from your own wallet

Pay-per-call endpoints now open a payment page instead of failing. You see the exact request being authorised before you approve it, paid calls link the transaction to the chain explorer, and a retried call reuses the payment you already made. Prices under a cent display properly instead of rounding away.

✋ Approvals and waiting steps

Agents can pause mid-run and wait for a person. A send-and-wait step holds the run, sends the request, and resumes exactly where it stopped once someone responds. Pending approvals show up in your MCP client.

▶️ Run forms from real examples

The run form is built from the agent's own example input — typed fields with descriptions instead of a raw JSON box. Agents that accept webhook calls publish that example on their listing, so you can see the expected shape before you buy.

💸 Publishing with markup

Add your margin on top of underlying costs when publishing, and see what you earn before you commit. Pricing is now one clear choice shaped by whether you bring your own keys, and the wizard says upfront whether an agent needs credentials from you.

🤖 More

Google Gemini is available as a model node. The full starter kit is granted on any first purchase, not just one path. Platform-held wallets are gone — indie.money no longer creates, holds or exports wallets for you, so bring your own.

🐛 Fixes

Share links survive titles with punctuation. Agent card run counts describe the window they actually cover. Card excerpts render cleanly when a description starts with a heading. Webhook payment prompts point at a reachable address, so the payment page loads every time. Catalog listings load faster. Payment settlement retries through provider slowdowns instead of failing the purchase.

One week of shipping. Here's everything that landed since the MCP Tools release.

🔌 64 New Integration Nodes

Your agents now work with 363 nodes — up from ~194. Highlights:

• Google — Drive, BigQuery, Books, Chat, Contacts, Cloud Storage, Translate, Tasks, Perspective, Cloud Natural Language, Business Profile, GSuite Admin
• Microsoft — SharePoint, Dynamics CRM, Graph Security
• SaaS — Elasticsearch, HelpScout, Onfleet, Splunk, SeaTable, KoBoToolbox, OpenWeatherMap, Sms77, TravisCi, HumanticAi
• AI Models — DeepSeek, xAI Grok chat models
• AI Tools — Think Tool, Vector Store Tool, JSON Document Loader
• Data — Summarize node for aggregation and statistics

Every node now links to its own help article on help.indie.money.

🔗 Share Agents

Share button on every agent page. Uses the Web Share API on mobile, falls back to clipboard copy on desktop. Copy Run ID button on the execution page too.

💰 Cheaper Agent-to-Agent Calls

When your agent calls another agent as a sub-step, billing is now smarter. Sub-workflow calls cost less and settle faster — no extra fees when both agents share the same budget.

🔑 Credential Improvements

• Boolean fields — toggle switches for on/off credential options
• Dropdown options — select from predefined values instead of free-text
• Conditional visibility — fields show/hide based on other field values
• URL validation — credential URLs validated before save
• OAuth scope fixes — LinkedIn, Microsoft Teams, Google Drive scopes corrected

🐛 Fixes

• Clipboard errors now show a toast instead of silently failing
• Sticky notes filtered from uploaded workflows — no more validation noise
• Wikipedia tool now sends proper User-Agent header
• HTTP Request node correctly splits JSON array responses into individual items
• AI sub-nodes show correct execution status (was stuck on "Executing")
• Webhook static data cleaned up on trigger deletion
• Upstream API error messages preserved instead of generic "500"

Still on Base Sepolia testnet. All transactions use test tokens — no real money involved. Keep breaking things. 🛠️