Sourcegraph Cody

Sourcegraph Cody

Sourcegraph Cody — Enterprise AI Coding Assistant with Cross-Repository Code Graph Context

English Introduction

Cody is the AI coding assistant built on top of Sourcegraph’s code intelligence layer. Instead of reading only the open file like GitHub Copilot or Cursor, Cody retrieves context from the entire indexed codebase — across hundreds of repositories — using Sourcegraph’s search API, SCIP symbol graph, and Deep Search. It answers “where is this API used?”, performs multi-file edits, explains legacy services, and applies Batch Changes across microservice fleets. IDE extensions cover VS Code, JetBrains (IntelliJ/PyCharm/WebStorm/GoLand), Visual Studio (preview), Neovim; also available in the Sourcegraph Web app and CLI. Enterprise deployment supports single-tenant cloud, self-hosted, and air-gapped with SAML/SSO, guardrails, and zero-training-on-customer-code. Not a standalone AI-first editor (≠ Cursor), not an autonomous SWE agent (≠ Devin/Factory), not a free solo-dev tool since July 2025 — it is the enterprise code-intelligence assistant for 100+ engineer orgs with monorepos and compliance needs. Sourcegraph’s sibling product Amp​ now absorbs the former free/pro agentic-use case.

Core Features

• Codebase-aware chat: multi-turn chat with @-mention for files/symbols/remote repos, default context = open file + current repo, expandable to 200+ repos via Sourcegraph index
• Autocomplete: whole-line and multi-line inline suggestions, local keyword + embeddings (Free historical) or full code-graph (Enterprise)
• Inline edits / Auto-edit: contextual diffs, multi-file refactors, explain-and-fix workflows
• Custom Prompts / Commands: reusable prompt templates for test-gen, review, doc-string insertion
• Deep Search (2026): natural-language codebase Q&A with shared team conversation history
• Advanced Code Graph context (Enterprise): SCIP symbols, cross-repo usage, dependency edges, Context Filters to exclude sensitive repos (auth/payments) from LLM context
• Model choice: Cody Gateway or BYO LLM — Claude Opus/Sonnet, GPT-4o/5, Gemini 1.5–3 Pro, DeepSeek-Coder (historical Free used DeepSeek-Coder-V2 + Claude 3.5 Sonnet)
• MCP Server: connect external tools, 20+ MCP endpoints
• Batch Changes: large-scale multi-repo PR generation (platform feature, Cody-assisted)
• Deploy: Sourcegraph Cloud / single-tenant / self-hosted / air-gapped; SAML/SSO, RBAC, audit logs, guardrails, HIPAA/SOC 2 Type II/ISO 27001/GDPR/FedRAMP-aligned
• No: no free tier (since Jul 2025), no standalone editor, no autonomous long-running agent (that’s Amp), no Jira/Linear native (gap vs GitHub Copilot Enterprise)

Pricing (2026-08, USD)

starting_price_for_new_buyer=$59/user/mo (Enterprise only); effective per-seat drops below $20/dev/moat100+devsonthe$16K platform contract. No monthly self-serve.

Access / Integration Methods

1. IDE extensions​ — VS Code, JetBrains suite, Visual Studio (experimental), Neovim
2. Sourcegraph Web app​ — cody in browser against indexed repos
3. Cody CLI​ — terminal chat + edits
4. Code hosts​ — GitHub, GitLab, Bitbucket (OAuth for Batch Changes), Perforce
5. MCP Server​ — external tool wiring (Slack, internal APIs)
6. Slack​ — multi-workspace + Enterprise Grid for chat-in-Slack queries
7. Admin​ — SAML/SSO, RBAC, Context Filters, Cody Gateway model routing
8. No​ REST “generate code” public API for indie devs; API/CLI exist for Enterprise automation only

Reviews (2026)

• Aggregate: G2 4.5/5 (74+ reviews, 68% 5-star), Gartner Peer Insights 4.4/5 (9), ProductHunt 4.7/5, Trustpilot 2.9/5 (few, billing complaints). ToolChase 4.3, AI Agent Index 3.8.
• Praised: unmatched cross-repo context, SCIP symbol accuracy, enterprise compliance (zero-retention, SOC2/ISO27001/HIPAA), model flexibility, VS Code/JetBrains parity, Batch Changes leverage.
• Complained: Free/Pro killed Jul 2025​ alienated solo devs; $59/seat steep under 25 devs; indexing config learning curve; occasional hallucination on complex multi-step logic; Visual Studio extension experimental; value locked behind Sourcegraph deployment; Amp migration unclear for old Pro users.
• Verdict: best-in-class for 100+ dev orgs with monorepos/regulated code; wrong for solo/small teams (→ Cursor $20,Copilot$10, Amp free, Aider/Continue).

Best For / Not For

Best for: enterprises with 100+ devs and 50+ repos; regulated industries (finance/healthcare/defense) needing air-gapped + SSO; teams already on Sourcegraph code search; microservice migrations across many services; security review of cross-repo secret/API usage.

Not for: solo developers, startups <10 devs, anyone needing a free tier, users wanting an AI-native editor (→ Cursor), autonomous ticket-solving agents (→ Devin/Factory/OpenHands), Jira-native PM workflows, non-GitHub/GitLab shops without admin bandwidth to index.

Competitors

• GitHub Copilot​ — $10–$39/user/mo, GitHub-native, file+issue context, no cross-repo code graph, no self-host
• Cursor​ — $20/mo Pro, AI-first VS Code fork, Composer agent strong, no SSO/audit/cross-repo intelligence
• Tabnine​ — autocomplete-first, privacy-focused, weaker codebase-wide chat
• Amazon Q Developer​ — free tier + AWS-native, weak cross-repo outside AWS
• Augment Code​ — enterprise code-graph rival, younger infra than Sourcegraph
• Windsurf / Codeium​ — $15–$40/mo, Cascade agent, lighter enterprise compliance
• Devin / Factory / OpenHands​ — autonomous SWE agents, different job than Cody’s assistant role
• Amp (Sourcegraph’s own)​ — free agentic CLI for ex-Cody-Free/Pro users, less enterprise context
• In-list overlap: none of your other cards (ElevenLabs/AIVA/PixVerse/Synthesia/Voicemod) touch Dev lane.