[FEATURE] Community feedback on `claude -p` pipe-mode bridge (TELAUDE) & contribution interest

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

I wanted to interact with Claude Code from Telegram while away from my terminal — send messages from my phone, get streamed responses back, manage multiple project sessions, and run scheduled tasks. Before Channels shipped, there was no official way to do this, so I built TELAUDE (https://github.com/UltraK18/TELAUDE), an open-source bridge that spawns claude -p as a child process and communicates via stdin/stdout.

Now that Channels is out as a research preview, I'd like feedback from the team on whether this claude -p pipe-mode approach is an acceptable pattern for third-party tooling — and whether there's interest in any form of collaboration or contribution.

Proposed Solution

This isn't a traditional feature request — it's more of a showcase + feedback ask. But if there's one concrete thing I'd propose:

Explicitly document whether claude -p pipe-mode is a supported integration pattern for community projects.

TELAUDE deliberately avoids the Agent SDK, OAuth token extraction, and subscription auth proxying. It calls the official CLI binary on the user's machine via pipe mode — the same way you'd use it in a terminal script. This means it inherits all native CLI features (session management, MCP integration, context compaction, tool permissions, prompt caching) without reimplementing or circumventing any of them.

I'd love to hear:

1. Is this approach something the team is comfortable with?
2. Does the team see value in community-built bridges like this — as complementary tooling, reference material, or something else?
3. I'm open to contributing in whatever way makes sense — official Channels plugins, documentation, edge-case patterns from months of battle-testing.
4. General thoughts, questions, or ideas are equally welcome.

Alternative Solutions

Before building TELAUDE, I considered:

• Agent SDK: Ruled out because Anthropic's ToS prohibits third-party use of subscription OAuth tokens with the SDK, and projects like OpenClaw/OpenCode/Cline were actively blocked for this.
• Waiting for official support: Channels didn't exist at the time, and the need was immediate.
• OpenClaw: Had security concerns and a looser permission model than I was comfortable with.

The claude -p child process approach was the most ToS-aligned path I could find.

Priority

Low - Nice to have

Feature Category

CLI commands and flags

Use Case Example

1. I'm working on multiple projects across different directories
2. I leave my desk but need to ask Claude to fix a bug or run a task
3. I send a message from Telegram → TELAUDE spawns claude -p --resume <sessionId> in the correct project directory
4. Claude streams its response back to Telegram in real time, with tool call visualization
5. When I'm back at my desk, the session is fully resumable in the terminal — same session ID, same context
6. For recurring tasks, I set up cron schedules that run autonomously with heartbeat health checks

Additional Context

• Repo: https://github.com/UltraK18/TELAUDE
• Architecture doc: https://github.com/UltraK18/TELAUDE/blob/master/ARCHITECTURE.md
• 196 commits, built with Bun + TypeScript + grammY + better-sqlite3
• Key features beyond Channels: multi-chapter sessions, per-message process spawning (no persistent terminal needed), cron scheduling, rich media support, built-in MCP server, TUI dashboard, OS-native encryption