ClawForge v2.0.0 — Fleet Forge Pivot

ClawForge is now an agent fleet forge for OpenClaw.

The pivot: From coding workflow CLI to agent fleet management. The name was always about this — ClawForge = forge Claws.

What's New

Fleet Management

• create — Interactive agent creation wizard with archetype templates
• list — Fleet overview with status indicators
• inspect — Deep view of agent DNA (config + workspace + bindings)
• edit / bind / clone / activate / deactivate / destroy — Full lifecycle
• migrate — Workspace isolation migration
• export / import — Share agents as .clawforge archives
• template — Manage reusable agent archetypes
• compat / upgrade-check — Fleet compatibility via clwatch
• doctor expanded with Fleet Health and Tool Versions

5 Built-in Archetypes
generalist, coder, monitor, researcher, communicator — with {{PLACEHOLDER}} substitution

clwatch Integration
Optional enrichment — model compatibility checks, deprecation monitoring, tool version tracking. Works without clwatch, smarter with it.

Changed

• sprint, review, swarm → use clawforge coding sprint|review|swarm
• All v1 commands still work (deprecation notice only)

Install / Upgrade

brew upgrade clawforge        # Homebrew
npm update -g @cyperx84/clawforge  # npm
uv tool upgrade clawforge     # uv

v1.6.3 — npm Binary Download on Install

npm Binary Download on Install

npm install -g @cyperx/clawforge now fully works — downloads the correct pre-built clawforge-dashboard and clawforge-web binaries for your platform automatically.

Supported platforms: darwin-arm64, darwin-amd64, linux-amd64, linux-arm64

If download fails (offline/firewall), install falls back gracefully and shows build-from-source instructions.

34/34 tests passing.

v1.6.2 — Emoji Forge Animation

Emoji Forge Animation

TUI startup animation is now a 10-frame emoji sequence:

⬛ cold → 🟠 embers → 🔥 fire → 🟥 hot forge → 🔨 hammer raised → 💥⚡ STRIKE → ✨ sparks → 💧 quench → 🤖 agents emerge → ✅ fleet online

140ms per frame, ~1.4s total.

v1.6.1 — quick-run CI Fix

CI Compatibility Fix

• quick-run --dry-run now works on CI environments without Claude/Codex installed
• Agent detection graceful during dry-run/help; real runs still validate and fail clearly
• 34/34 tests passing

v1.6.0 — quick-run: Direct Agent Execution

quick-run: Zero-Overhead Direct Agent Execution

New command for running an agent directly without the full sprint workflow ceremony.

clawforge quick-run "Explain what this codebase does"
clawforge quick-run "Fix the failing test" --dir ~/github/myrepo
clawforge quick-run "Summarize changes" --no-track --save /tmp/out.log

• No worktree, no branch, no tmux
• Streams output live to terminal
• Logs captured to ~/.clawforge/registry/logs/
• Tracked in clawforge status (unless --no-track)

34 test suites, all passing.

v1.5.5 — TUI Observability: Logs + Diff Keys

TUI Observability: Logs + Diff Keys

Fixes the "can't see what the agent did" problem when sessions are headless or already finished.

New keys

• l — show last 50 lines of agent log in preview pane
• d — show git diff --stat HEAD for agent's worktree
• Enter — attaches if session is live; shows log if session ended
• Esc — close preview/log panel

Log capture

• Agent output now teed to ~/.clawforge/registry/logs/<branch>.log automatically
• log_path stored in registry and surfaced to TUI

Full suite: 33/33

v1.5.4 — TUI Attach + Session Persistence

TUI Attach + Session Persistence

• Keep agent tmux sessions alive after command completion so Enter attach still works.
• Improve Enter behavior when target session is missing (graceful fallback instead of silent failure).
• Add verbose mode for Claude launches to make attached output clearer.

This fixes the UX where Enter looked broken when agents ran in fast/headless mode.

v1.5.3 — TUI Fixes

TUI Fixes

• Init race fixed: Agents load instantly on startup (no more 2s "No agents found" flash)
• False failure removed: Running tasks stay running even when tmux session hasn't spawned yet
• Unicode-safe columns: Emoji in status/descriptions no longer misalign the table

Full suite: 33/33 passing

v1.5.2 — Test Harness Stability

Test Harness Stability

• Replaced fragile echo | grep -q assertions with here-string checks across all 29 test files.
• Eliminates intermittent broken-pipe failures under set -o pipefail on GitHub Actions macOS runners.
• All 33 test suites passing.

v1.5.1 — CI Test Flake Fix

CI Test Flake Fix

• Fixed flaky test-dashboard behavior on GitHub Actions macOS runners.
• Replaced echo | grep -q in assertion helper with deterministic here-string matching.
• Keeps full dashboard test coverage while removing intermittent broken-pipe failures under set -o pipefail.

Validation

• Full suite passing: 33/33