# FlyMyAI Headless Cloud - platform log

Current version: **1.10.10** (2026-10-07). Newest first; this page shows the latest 10 entries.

- What the platform does today (the current snapshot): https://flymy.ai/skill.md
- Every entry, machine-readable: https://flymy.ai/changelog.json
- Agents: call `whats_new` on the Agents MCP with `since` set to the last version you saw.

## 1.10.10 - 2026-10-07 - Nano Banana 2.1: Google's newest image model, up to 4K

- **models** - Nano Banana 2.1 by Google makes and edits images up to 4K from a prompt and up to 14 reference images, with readable text and characters that keep their look across edits. You pay Google's token price: $0.0336 per 1K image plus thinking tokens, about $0.04 a run. Use: `recommend_model`, `get_pricing`, `run_model`, `/api/v1/flymyai/nano-banana-2_1/predict`. [Docs](https://docs.flymy.ai/guides/nano-banana-2-1)
  Try: "Make a 2K poster for a jazz night in Lisbon titled "Noite de Jazz" with Nano Banana 2.1."

## 1.10.9 - 2026-10-07 - Leads use their subagents' files in code; Instagram in Agent Teams at the provider's price

- **agents** - A lead agent sees and reads the files its subagents save and can pass them straight into its code run: the code finds them under /tmp/workspace/inputs, so collected data is processed without copying it by hand. Use: `run_agent`, `get_run`. [Docs](https://flymy.ai/skill.md)
  Try: "Have a subagent save a CSV of sales by region, then total it in code and tell me the sum."
- **tools** - Instagram data in Agent Teams on FlyMyAI's shared account costs the same as in a single agent: $0.02 per request, what the data provider charges. With your own HikerAPI key you pay only compute time. Use: `run_agent`. [Docs](https://docs.flymy.ai/mcp/instagram-public)

## 1.10.8 - 2026-10-07 - Subagents hand back what they have, and leads finish with partial results

- **agents** - A subagent that reaches its budget or keeps repeating the same call is stopped with its saved result and the reason, and its lead finishes with what the subagents delivered instead of redoing their work or waiting. Use: `run_agent`, `get_run`, `get_team_run`. [Docs](https://flymy.ai/skill.md)
  Try: "Have three subagents each summarize one of these three articles, and give me what they finish."
- **agents** - An API, frozen or scheduled run that cannot do part of its task, such as reading a file that does not exist, finishes as completed and names the missing part instead of failing. Use: `run_agent`, `run_frozen`, `schedule_agent`. [Docs](https://flymy.ai/skill.md)

## 1.10.7 - 2026-10-07 - Pick the model your project's agent runs on

- **app** - Choose the model of a project's agent: the Model list in the project panel, `model` in projects_agent, the REST endpoint or the Python SDK 1.2.0rc11. A new project agent starts on the strongest model, and your choice stays through platform updates. Use: `projects_agent`, `list_agent_models`, `/api/v1/agents/projects/{id}/agent/`. [Docs](https://docs.flymy.ai/agents/guides/projects#pick-the-project-agents-model)
  Try: "Switch my project's agent to GPT-5.6 Terra."

## 1.10.6 - 2026-10-07 - Instagram at the provider's price, project spend with its agent, vision refusals say why

- **tools** - Instagram data on FlyMyAI's shared account costs what the data provider charges for each request: $0.02 per request, and the balance check is free. With your own HikerAPI key you pay only compute time. Use: `search_tools`, `execute_tool`. [Docs](https://docs.flymy.ai/mcp/instagram-public)
  Try: "Get the follower counts of @nasa and @spacex."
- **app** - A project's spend now includes its project agent: the model and tool calls of its chats and of the subagents it starts count next to page calls and servers. Use: `projects_get`, `projects_list`. [Docs](https://docs.flymy.ai/agents/guides/projects)
  Try: "How much did my project spend this week, including its agent's chats?"
- **models** - When a vision model refuses a request, the answer says why in a reason field next to the usual error, for example a prompt longer than 100,000 characters. Use: `analyze_media`, `/api/v1/flymyai/gemini-3_8-flash-vlm/predict`. [Docs](https://docs.flymy.ai/api/models-endpoints)

## 1.10.5 - 2026-10-06 - Share links open straight to the page, and your own version is one prompt away

- **artifacts** - A shared artifact link now opens the page under one thin bar. With sources shared, Make your own version gives copy-ready steps for Claude Code, Codex and chat apps: connect FlyMy.AI, paste one prompt, and your agent clones it and publishes your version. Agents fetching the link see them too. Use: `clone_artifact`, `read_artifact_files`, `publish_artifact_version`, `/api/v1/artifacts/shared/{handle}/`. [Docs](https://docs.flymy.ai/agents/guides/artifacts#continue-from-a-share-link)
  Try: "Make my own version of the FlyMy.AI artifact my colleague shared with me and change its title."

## 1.10.4 - 2026-10-06 - Python SDK: add your own MCP servers with OAuth sign-in

- **sdk** - Python SDK 1.2.0rc9 (pip install 'flymyai==1.2.0rc9') adds client.mcp_servers: add an MCP server by URL, sign in to an OAuth server with a link a person approves, then call its tools from your code. Tokens and the client secret never come back. Use: `/api/v1/agents/mcp-servers/`, `/api/v1/agents/mcp-servers/{id}/authorize/`. [Docs](https://docs.flymy.ai/mcp/custom/connect-custom-server#from-an-assistant-or-from-code)
  Try: "Write a Python script with the FlyMyAI SDK that adds Linear's MCP server with OAuth and prints the sign-in link."

## 1.10.3 - 2026-10-06 - Slack as you: search, DMs, files, canvases and lists

- **tools** - Connect Slack once and agents work as you: search messages and files, read and post in channels and direct messages, share files, write canvases and lists, and set reminders. The bot connector is now called Slack Bot and posts as the FlyMyAI app. Use: `search_tools`, `execute_tool`. [Docs](https://docs.flymy.ai/mcp/slack)
  Try: "Search my Slack for messages about the release this week and summarize what was decided."

## 1.10.2 - 2026-10-06 - Wan 3.0 video with sound, up to 30 seconds

- **models** - Wan 3.0 by Alibaba makes video with sound from a prompt, from a first and last frame, or from reference images, clips and audio: 2-30 seconds or a length it picks, at 480P, 720P or 1080P. You pay per generated second, $0.05 to $0.20 by resolution. Use: `recommend_model`, `get_pricing`, `run_model`, `/api/v1/flymyai/wan-3_0-t2v/predict/async/`. [Docs](https://docs.flymy.ai/guides/wan-3-0)
  Try: "Make a 10-second 720P video with sound of a red fox running through fresh snow at dawn with Wan 3.0."

## 1.10.1 - 2026-10-06 - Grok 4.7 with private mode

- **models** - Grok 4.7 by xAI answers prompts and chats with text and images through /predict, the OpenAI-compatible API and run_model. Choose how hard it reasons, from low to xhigh, and turn on private mode to keep the prompt and the answer out of FlyMyAI storage. Use: `run_model`, `get_pricing`, `/api/v1/flymyai/grok-4_7/openai/v1/chat/completions`, `/api/v1/flymyai/grok-4_7/predict`. [Docs](https://docs.flymy.ai/guides/grok-4-7)
  Try: "Ask Grok 4.7 in private mode to rewrite my launch email in three tones."

25 older entries: https://flymy.ai/changelog.json