Telegram Channel + Smarter Agent Creation

Kindship now has an official Telegram channel, a smarter agent-creation flow, and more reliable planning chat when tools are involved.

Telegram Channel: Chat With Your Agents Anywhere

If you prefer messaging over browser tabs, you can now use Telegram as a first-class way to talk to your agents.

Chat with your agent - Send a message and get responses just like the web app
Switch between agents - Use /agent to pick a different agent from an inline menu
Check your current agent - Use /status to confirm which agent is active
Get help fast - Use /help to see available commands
Get proactive updates - Agents can message you first (for example, when something is ready), with optional auto-switching to the relevant agent
Unlink any time - Use /logout to disconnect Telegram from your Kindship account

Private chats only - Group and channel messages are ignored to avoid accidental leakage
One-time linking tokens - Linking uses short-lived, single-use tokens (no permanent secrets in chat)
Verified webhooks - Telegram updates are verified before processing

Creating an agent now includes a dedicated naming step that helps you get to a strong name and a clear prime directive faster.

Better starting point - Generate name suggestions instead of staring at an empty input
Duplicate detection - Avoid colliding with existing agent names
Expanded blacklist - Steer clear of awkward, confusing, or low-signal names
Prime directive included - Get a suggested directive you can keep as-is or edit before creating the agent

The result is an agent that feels intentional from day one, and a directive that is easier for the agent to follow consistently.

New agents can now automatically provision their GitHub documents repository during creation, so document sync is ready without extra setup.

Faster first write - Start creating documents immediately, without a separate provisioning step
Safer storage by default - Documents are backed by GitHub instead of living only inside a container
Smoother onboarding - Fewer buttons and fewer "now go configure X" moments

Planning chat now does a better job of showing what the agent is doing when tools are involved.

Tool execution visibility - You can see when the agent is calling tools and what it produced
Multi-turn tool flows - Tool-driven work can run across multiple turns without breaking streaming
Cleaner context - User messages are included in history so the agent has the right context for follow-ups
Fewer broken links - Relative URLs produced during tool execution render correctly
Stronger default reasoning - Planning chat now uses a more capable default model for higher-quality plans and more consistent follow-through

For team accounts, planning chat conversation tracking and runs are now more reliably scoped to the right account.

Correct account association - Conversation runs and related execution data stay attached to the owning account
Stronger access checks - Additional verification prevents cross-agent access edge-cases

If you run agents as a team, this reduces "why can't I see this run?" confusion and makes audit and debugging simpler.