Migrate from Sourcegraph

Sourcegraph discontinued the individual/indie tier in July 2025. If you were using Sourcegraph for code search and navigation on a self-hosted or cloud account, this guide walks through replicating that workflow with Canopy.

What changes

The biggest practical difference: Canopy runs entirely locally. There’s no server to run, no Docker container to maintain, no cloud account required. The index lives in .canopy/index/ inside your repo and is updated incrementally as files change.

Feature mapping:

Sourcegraph	Canopy equivalent
Code search (literal)	`canopy search <query>` / `canopy_search`
Code search (regex)	Use `canopy pattern` for AST-aware structural matching, or pipe `canopy search` output to `grep -E '<pattern>'` for regex post-filtering. The CLI does not currently expose a `--regex` flag.
Structural search	`canopy pattern <ast-grep pattern>` / `canopy_pattern_search`
”Find references”	`canopy_trace_dependents` (MCP tool — no CLI peer)
“Go to definition”	`canopy_search_symbols <symbol>` (MCP tool — no CLI peer)
Cross-repo search	Index each repo separately; search each with separate MCP server
Code navigation in browser	Use with Claude Code, Cursor, or Zed for in-editor navigation
Batch changes	Not available — Canopy is a read-only intelligence tool
Insights/metrics	`canopy status` for index counts; `canopy_architecture_map` (MCP) for structural metrics

Step 1: Install Canopy

curl -fsSL $CANOPY_DOWNLOAD_URL/releases/latest/canopy-linux-x86_64 \
  -o canopy
chmod +x canopy
sudo mv canopy /usr/local/bin/canopy
canopy --version

See Install Canopy for all platforms and Gatekeeper handling on macOS.

Step 2: Index your repo

cd /path/to/your/repo
canopy index . --with-search --with-git

--with-search enables full-text search (the Sourcegraph equivalent). --with-git adds git history to canopy_prepare results.

For large repos (>10K files), indexing takes 30–120 seconds. Subsequent incremental runs are much faster — typically under 5 seconds.

Step 3: Connect to your editor

If you used Sourcegraph’s browser-based code navigation, the closest equivalent is connecting Canopy to your editor via MCP:

Once connected, you can ask your AI assistant to search and navigate using Canopy tools directly.

Step 4: Replicate your searches

Literal search:

Sourcegraph: repo:^github\.com/myorg/myrepo$ checkout

Canopy CLI:

canopy search "checkout"

Canopy MCP (in editor):

Use canopy_search with query="checkout"

Regex search:

Sourcegraph: repo:^github\.com/myorg/myrepo$ /createPayment$.*$/

Canopy doesn’t expose a --regex flag on canopy search. For pattern-based matching, prefer canopy pattern (ast-grep over the AST — handles all the semantically-equivalent forms regex would have to enumerate). For raw regex over indexed text, pipe through grep -E:

canopy search "createPayment" | grep -E "createPayment\(.*\)"

For function-call shape matching that’s more robust than regex:

canopy pattern "createPayment($$$)" --language typescript

Structural search:

Sourcegraph structural: console.log(...)

Canopy pattern (ast-grep syntax):

canopy pattern "console.log($$$)" --language javascript

Find all references to a function:

Sourcegraph: “Find references” on a symbol in the UI

Canopy MCP:

Use canopy_trace_dependents on src/lib/payments.ts

There is no CLI peer for “find references” — canopy_trace_dependents is exposed only through MCP. From the editor, ask your AI assistant to call it; from the shell, the closest read-only check is grepping import statements:

# Approximate: list files that import the module by string match.
# Misses re-export chains and dynamic imports — use canopy_trace_dependents
# from your editor for the structural answer.
canopy search "from '../lib/payments'"

Activate a license (optional)

Canopy Community tier supports 1 repo and basic CLI tools — sufficient for personal projects. For multi-repo support and the full 31 MCP tools, activate a license:

canopy activate <your-license-key>

A 14-day free trial is available at canopy.gulfshieldtech.com. Card required upfront; no charge until day 15.

What Canopy doesn’t have

Be aware of these gaps before migrating:

Cross-repo search in a single query — Canopy indexes each repo separately. You can’t search across 50 repos with one query the way Sourcegraph Enterprise supported. Workaround: use a monorepo index if your repos are related.
Web UI — Canopy has no browser interface. Results come through the CLI or your editor’s AI assistant.
Batch changes — Canopy is a read-only intelligence tool. It doesn’t write code.
Saved searches / code monitors — No persistent search subscriptions. Run searches on demand or in CI.

Common pitfalls

Search returns too many results Sourcegraph defaulted to repo-scoped search. Canopy searches the entire indexed tree. Use --language to filter by language and --limit to cap the result count — there is no --include glob flag on the CLI:

canopy search "checkout" --language typescript --limit 20

For path-prefix scoping, canopy_search (the MCP tool) accepts path_prefix=src/ — call it from your editor when you need that filter.

“Index not found” after indexing Canopy stores the index relative to the path passed to canopy index. If you indexed /home/you/repos/myapp, start canopy serve with the same path. Running canopy serve . from a different working directory won’t find the index.