docs: refresh README positioning and getting started#168
Conversation
edwin-zvs
left a comment
edwin-zvs
left a comment
There was a problem hiding this comment.
first round of review
|
Addressed review feedback in 74e5b85:
Validation:
|
| Bare names auto-detect: `gpt-*` / `o[1-5]*` → OpenAI, `claude-*` → | ||
| Anthropic, anything else → Ollama. When in doubt, use the explicit | ||
| prefix. | ||
| ## Smoke test |
There was a problem hiding this comment.
Let's remove Smoke test section from README.md
| be loaded, or whose harness rejects the respawn are marked `Errored` | ||
| (transcript + scrollback remain readable). | ||
|
|
||
| Deferred to later milestones: |
There was a problem hiding this comment.
Let's remove this section
| @@ -0,0 +1,146 @@ | |||
| # Harnesses and session modes | |||
|
|
|||
| ## Milestone 1 scope | |||
There was a problem hiding this comment.
let's remove this section
| @@ -0,0 +1,32 @@ | |||
| # Agent-controlled agentd (MCP) | |||
There was a problem hiding this comment.
Can we name it 'Unified tool layer'? Also name it in this way from README
| - Passes `--mcp-config <path>` to the child | ||
| - Sets `AGENTD_SESSION_ID=<session_id>` in the child's environment | ||
|
|
||
| The MCP server exposes a read + write tool surface that mirrors the IPC: |
There was a problem hiding this comment.
list of mcp server and tools organized in sections and table, rather than statements
| # zarvis built-in agent | ||
|
|
||
|
|
||
| `zarvis` is a self-contained headless agent that ships with agentd. It |
There was a problem hiding this comment.
remove ('headless'). Also mention "many PRs for agentd repository has been made from 'zarvis'"
edwin-zvs
left a comment
edwin-zvs
left a comment
There was a problem hiding this comment.
left more comments
|
Addressed the latest review feedback in b6438a7:
Validation:
|
|
Added remote-control docs in 59d4f4b:
Validation: . |
|
Added remote-control docs in 59d4f4b:
Validation: |
| output, spawn helpers, send input, inspect diffs, and drive Chrome. | ||
| - **Remote control when you step away** — `/remote-control` opens a | ||
| browser-accessible web client with a QR code, basic auth, and a local modeline | ||
| badge when remote clients are attached. |
There was a problem hiding this comment.
Remove 'basic auth, and a local ....'. Instead "Connect from your phone, no service signup, no setup required"
|
Addressed the latest README wording comment in 9d TBD after push: the remote-control bullet now says it opens a browser-accessible client with a QR code and emphasizes connecting from your phone with no service signup or setup required. Validation: |
|
Tuned the README tone in b985fec to hint at a more tactile, alive terminal experience without naming specific visual effects:
Validation: |
| # agentd | ||
|
|
||
| A terminal **agent fleet** — run and supervise multiple coding-agent sessions across heterogeneous harnesses (Claude Code, Codex, generic shell, ...) from one TUI. | ||
| **Command a fleet of coding agents without losing the thread.** |
There was a problem hiding this comment.
"Command a feet of agents, designed for hackers"
| Launch, watch, steer, and compose multiple AI coding sessions in a workspace | ||
| that feels alive on the wire. Keep Claude Code, Codex, Antigravity, Zarvis, shell | ||
| tasks, and future harnesses moving together instead of juggling tabs, transcripts, | ||
| and background processes. |
There was a problem hiding this comment.
Create Codex, Claude Code, Antigravity, Zarvis sessions all in one place. Or let your agent coordinate them on terminal that is crafted hardcore hackers like you. Remote control from your phone when you're in motion.
|
|
||
| Status: **early — M2 (PTY mode) just landed. Wire protocols may still break.** | ||
| Create Codex, Claude Code, Antigravity, and Zarvis sessions all in one place. Or | ||
| let your agent coordinate them in a workspace crafted for hardcore hackers like |
There was a problem hiding this comment.
in this particular case, let's use word 'terminal'.
"in a workspace crafted for" -> "in a terminal"
1 participant
Summary
Testing
No screenshot/video: documentation-only change.