Watchfire vs Sourcegraph Cody (and Amp).
Sourcegraph's coding assistants — Cody (chat with deep codebase context) and Amp (agentic coder) — backed by Sourcegraph's code intelligence platform and indexed cross-repo search.
The verdict
Pick Sourcegraph Cody or Amp when your team already runs Sourcegraph and you want agent answers grounded in indexed, cross-repo context. Pick Watchfire when you want local-first, task-shaped work with the agent CLI you already use.
At a glance.
The axes that actually differ between the two tools. The prose below this table is where the nuance lives — don’t pick a tool from a row alone.
| Axis | Sourcegraph Cody (and Amp) | Watchfire |
|---|---|---|
| Agent backend | Sourcegraph-managed models (mostly Anthropic and OpenAI), with enterprise bring-your-own-key options. | Pluggable CLI: Claude Code, Codex, opencode, Gemini, Copilot CLI, or Cursor Agent. |
| Isolation | Cody runs as an IDE extension editing your working tree; Amp runs an agent that proposes edits and PRs through Sourcegraph. | Per-task git worktree on a watchfire/<n> branch on your machine. |
| Sandbox | Editor-scoped for Cody; Amp's runtime is Sourcegraph-managed. | Seatbelt on macOS, Landlock or Bubblewrap on Linux; sensitive paths blocked by default. |
| Parallelism | One Cody conversation per IDE window; Amp parallelism is plan-dependent. | One agent per project, many projects concurrently from a single daemon. |
| Git workflow | Cody edits the working tree; Amp opens PRs through Sourcegraph. | Reviewable branch per task; auto-merge is opt-in. |
| Autonomy | Cody is chat-and-edit; Amp is the autonomous tier. | Interactive by default; autonomous opt-in via Wildfire mode. |
| Local vs cloud | Hybrid — the extension runs locally, but retrieval and inference go through Sourcegraph's cloud or a self-hosted Sourcegraph instance. | Local — nothing leaves your machine unless you ship it. |
| Multi-project | Codebase-aware via the Sourcegraph index across many repos. | First-class — many projects active concurrently. |
| Open source | Cody had an open-source extension; Amp and the Sourcegraph backend are closed-source enterprise products. | Apache-2.0. |
| Pricing | Free tier for individual Cody use; Team and Enterprise tiers are seat-priced; Amp is paid. | Free; you pay model API costs of the CLI you pick. |
Agent backend
Sourcegraph Cody (and Amp)
Sourcegraph-managed models (mostly Anthropic and OpenAI), with enterprise bring-your-own-key options.
Watchfire
Pluggable CLI: Claude Code, Codex, opencode, Gemini, Copilot CLI, or Cursor Agent.
Isolation
Sourcegraph Cody (and Amp)
Cody runs as an IDE extension editing your working tree; Amp runs an agent that proposes edits and PRs through Sourcegraph.
Watchfire
Per-task git worktree on a
watchfire/<n>branch on your machine.Sandbox
Sourcegraph Cody (and Amp)
Editor-scoped for Cody; Amp's runtime is Sourcegraph-managed.
Watchfire
Seatbelt on macOS, Landlock or Bubblewrap on Linux; sensitive paths blocked by default.
Parallelism
Sourcegraph Cody (and Amp)
One Cody conversation per IDE window; Amp parallelism is plan-dependent.
Watchfire
One agent per project, many projects concurrently from a single daemon.
Git workflow
Sourcegraph Cody (and Amp)
Cody edits the working tree; Amp opens PRs through Sourcegraph.
Watchfire
Reviewable branch per task; auto-merge is opt-in.
Autonomy
Sourcegraph Cody (and Amp)
Cody is chat-and-edit; Amp is the autonomous tier.
Watchfire
Interactive by default; autonomous opt-in via Wildfire mode.
Local vs cloud
Sourcegraph Cody (and Amp)
Hybrid — the extension runs locally, but retrieval and inference go through Sourcegraph's cloud or a self-hosted Sourcegraph instance.
Watchfire
Local — nothing leaves your machine unless you ship it.
Multi-project
Sourcegraph Cody (and Amp)
Codebase-aware via the Sourcegraph index across many repos.
Watchfire
First-class — many projects active concurrently.
Open source
Sourcegraph Cody (and Amp)
Cody had an open-source extension; Amp and the Sourcegraph backend are closed-source enterprise products.
Watchfire
Apache-2.0.
Pricing
Sourcegraph Cody (and Amp)
Free tier for individual Cody use; Team and Enterprise tiers are seat-priced; Amp is paid.
Watchfire
Free; you pay model API costs of the CLI you pick.
When to pick Sourcegraph Cody (and Amp).
Pick Sourcegraph Cody or Amp when your team already runs Sourcegraph and you want every answer grounded in indexed, cross-repo code intelligence — not just the files in the current workspace. That cross-repo context is genuinely hard to replicate: Cody and Amp can pull a struct definition from one repo, a caller from another, and a config from a third into the same prompt. For enterprise teams that already pay for Sourcegraph for code search and review, layering Cody and Amp on top is a low-friction way to get agentic coding with consistent context across the whole codebase. The enterprise tier also brings audit logs, SSO, and admin controls that Watchfire does not ship.
When to pick Watchfire.
Pick Watchfire when you want local-first, task-shaped work without a Sourcegraph dependency and without a vendor in the inference path. Watchfire runs on your laptop or workstation; the daemon, worktrees, and sandboxes all live locally, and the agent CLI you pick — Claude Code, Codex, opencode, Gemini, Copilot CLI, or Cursor Agent — decides what context to pull from the repo in front of it. There is no central index to maintain, no seat to provision, and no inference to route through someone else's cloud. If your repo is the scope and you would rather pay per-token to your model provider than per-seat to a platform, Watchfire is the fit.
Can they coexist?
Yes — they target adjacent problems and a team can use both. Cody and Amp are at their best when grounded in Sourcegraph's index across many repos; Watchfire is at its best when the unit of work is a single task in a single project with strong local isolation. A pragmatic team setup is to use Cody for cross-repo chat and code intelligence and Watchfire to actually run the long, isolated task runs that change code. They do not write to the same place: Cody edits the IDE's working tree, Watchfire works inside .watchfire/worktrees/ on its own branch.
Related reading.
- ArchitectureDaemon, clients, and where data lives.
- Worktree isolationWhy each task runs on its own branch.
- Agent modesChat, Task, Start All, Wildfire — when to use which.
- See the full comparison matrixEvery adjacent tool, side by side.
- Start with a ready taskCopy-paste starter task YAMLs for the most common work.
Get started with Watchfire
Install in seconds. Define tasks. Let agents ship code for you.
Download for macOSAvailable for macOS, Linux, and Windows
# Install via Homebrew (macOS)brew tap watchfire-io/tapbrew install --cask watchfire-io/tap/watchfire# Set up your project and gowatchfire initwatchfire task add "Build the login page"watchfire start --allIncludes GUI, CLI, and daemon. Also available via Homebrew.