Vercel and Slack are also entering the coding tools market.

fx — a tiny native code agent
https://fx.sh/
https://github.com/vercel-labs/fx
fx — Vercel Labs' answer to bloated agentic TUIs — is for now an experimental CLI harness written in Zig: a 6–8 MB binary, microsecond startup, minimal system prompt, and a UI closer to a Unix shell than an "IDE in the terminal." Licensed under Apache-2.0. Includes a Wasm build (live demo directly in the browser), skills, MCP, subagents, and ACP for editors.

Models are routed via Vercel AI Gateway or via Codex (fx login codex) or Grok (fx login grok) subscriptions; the current default in the demo is glm-5.2. Currently at v0.0.5 — "use at your own risk."

Slack Code: Agentic Coding for Teams
https://www.salesforce.com/introducing-slack-code/
Salesforce is launching Slack Code — dedicated code channels inside Slack where you can tag a code agent (Claude, Devin, GitHub Copilot, ChatGPT, Vercel agents) and collaboratively review diffs, previews, and approve work without leaving chat. It features an Agents tab and "Add to Slack" for agents from Lovable, n8n, LangChain, and others. At launch, it works with partner agents and is available on any Slack plan — but requires separate access to each agent.

The idea is straightforward: currently, the human-agent conversation happens in a private tab, the team doesn't see it, and context is lost. In Slack Code, a channel is created per task and archived when no longer needed, with history preserved as a log. They demo a workflow where a PM tags an agent on a bug directly, and an engineer only needs to review the diff and approve the PR.

Proliferate: Open-Source IDE for Multiple Agents
https://github.com/proliferate-ai/proliferate/releases
https://proliferate.com/
Proliferate is an attempt to assemble a team of agents with branch isolation and shared review. The project is open-source (AGPL-3.0) and runs multiple native harnesses in parallel within a single workspace: Claude Code, Codex, OpenCode, Cursor, Grok. Each task is a separate git worktree (or a cloud sandbox that stays alive when the laptop is closed) with subagents, shared MCP/skills, and beta workflows (nightly reviews, CI triage, dependency bumps).

jcode for Concurrent Agent Collaboration
https://jcode.sh/
https://github.com/1jehuang/jcode
https://www.ycombinator.com/companies/jcode
jcode is an open-source (MIT) terminal agent written in Rust by Jeremy Huang, Y Combinator. The project's mission is to parallelize dozens of tasks simultaneously. Its key differentiator is ultra-low memory overhead: per project benchmarks, 10.4 MB per additional session vs. 212.7 MB in Claude Code — so 10 jcode sessions consume less than half the memory of a single Claude Code instance.

The TUI remains fully featured with support for images, LaTeX/math, diagrams/Mermaid, live progress for background tasks, swarm mode with file-conflict notifications, and inter-agent messaging.

Orca by Stably
https://onorca.dev/
https://github.com/stablyai/orca
Orca is a standalone open-source ADE (Agent Development Environment, MIT). It's a desktop IDE for parallel agents: each task gets its own git worktree + dedicated terminal (Ghostty-inspired) + browser — you can run Claude Code, Codex, OpenCode, Cursor side-by-side, with mobile companions and Orca CLI included.

https://www.youtube.com/watch?v=vZUUHMCBoQg

Practical value comes from the Workspace Board for distributing tasks across agents and a unified Diff Review for reviewing/merging without jumping between branches, plus built-in Chromium with Design Mode (DOM inspection → copy markup to agent for targeted layout fixes), tab splits, integrated editor, and git-workflow automation (worktree creation via Cmd+N + setup script pnpm install, staging/commits/PRs).

#agenticcoding #cli #harness #multiagents