Sourcegraph Problems & Issues
Real user-reported problems, complaints, and limitations for Sourcegraph. Based on analysis of 12 signals.
Overall score: 55/100
Issues by Category
Occasional hallucinations in AI output
Like all LLM-based tools, Cody experiences occasional hallucinations. Generated code may look correct but contain subtle bugs. Requires careful human oversight and validation. Not fully trustworthy for critical code paths.
Confusing product transitions (Cody to Amp)
Individual users must migrate from Cody to Amp - new product with different pricing and features. Transition credits provided but workflow disruption significant. Enterprise customers unaffected but product direction feels unstable.
AI struggles with complex multi-step logic
Cody struggles with complex, multi-step logic and long dependency chains. Experiences occasional hallucinations requiring human oversight. Context windows are finite - large repositories require iterative prompting for accuracy. Not suitable for fully autonomous complex tasks.
Over-reliance risk for newer developers
For newer developers, Cody can become a crutch if outputs aren't validated. AI-generated code without understanding is a risk. Reviews warn against dependency on AI for learning - fundamentals may be skipped.
Cody Pro discontinued July 2025
Cody Free, Cody Pro, and Cody in Starter plans were deprecated July 23, 2025. Individual users forced to migrate to Amp (new product) or find alternatives. $10-40 in transition credits offered but product direction unclear. Major disruption for existing users.
Enterprise pricing very expensive
Enterprise tier costs $59/user/month - one of the most expensive AI coding tools. For 100 developers, that's $70,800/year. Starter at $19/user/month limited to 50 users. Volume discounts require sales negotiation. Cost prohibitive for many teams.
Pricing based on lines of code in some cases
In certain situations, pricing is based on combination of active users AND lines of code indexed. Large monorepos may face higher costs than expected from per-seat pricing alone. Requires sales conversation to understand true costs.
Account bans for minor violations
User reported being banned for accidentally creating a second account by logging in with GitHub after paying subscription. Also experienced 429 rate limiting on paid account. When contacting support as recommended, account was banned for 'Acceptable Use Policy' violation. Heavy-handed enforcement.
Individual users deprioritized
Product direction clearly favors enterprise customers. Individual plans discontinued. Free/Pro users pushed to new Amp product. Company explicitly 'simplifying offerings to better serve customers' - meaning enterprise, not individuals.
Company instability - multiple layoffs
Sourcegraph experienced multiple layoffs in 2025, including 8% workforce reduction (24 employees) and multiple marketing team cuts within 6 months. Company went from growth to survival mode. Employee reviews describe leadership as 'toxic' and product as 'directionless.'
Viewed as 'vitamin' not 'painkiller'
Customer reviews indicate Sourcegraph was seen as a 'vitamin' (nice-to-have) rather than 'painkiller' (essential). Hard to justify ROI to management. Code search valuable but not critical enough to survive budget cuts at many organizations.
Large repositories require iterative prompting
Context windows limit how much code Cody can understand at once. Extremely large or tangled repositories may require multiple prompts to get accurate results. Not as seamless as marketing suggests for enterprise-scale codebases.
Regret Patterns
Common reasons users regret choosing Sourcegraph.
Cody Pro discontinued without alternative
Paid for Cody Pro expecting continued AI assistance. Product discontinued July 2025 with migration to Amp required. Transition credits insufficient for long-term use. Had to find new tool and rebuild workflow.
Enterprise costs exceeded budget expectations
Evaluated Sourcegraph for team assuming reasonable pricing. Discovered $59/user/month - 3x Copilot. 100 developers = $70k/year. Could not justify ROI to management. Had to choose cheaper alternative.
Company instability raised vendor risk concerns
Deployed Sourcegraph for enterprise code search. News of layoffs, product discontinuation, and negative employee reviews raised red flags. Management questioned long-term vendor viability. Considering migration despite sunk costs.
Viewed as 'nice-to-have' in budget cuts
Implemented Sourcegraph for code search benefits. During budget review, product categorized as 'vitamin not painkiller.' Hard to prove direct ROI. Lost tool during cost-cutting despite team finding it useful.
Account banned for minor policy violation
Created second account accidentally by logging in with GitHub after paying subscription. Account banned for 'Acceptable Use Policy' violation. Support unhelpful. Lost paid subscription and access.
AI limitations became apparent at scale
Expected Cody to handle complex enterprise codebase seamlessly. Found context window limitations, hallucinations, and need for iterative prompting on large repos. Marketing overpromised actual capabilities.
When Sourcegraph Breaks Down
Specific scenarios where Sourcegraph fails or underperforms.
Individual plan discontinued
Cody Free and Pro ended July 2025. Individual users must migrate to Amp or find alternatives. No direct replacement offered. Workflow disruption for all non-enterprise users.
Enterprise costs during budget review
At $59/user/month, Sourcegraph is hard to justify in cost-conscious environments. Seen as 'vitamin' not 'painkiller.' Often first to be cut when budgets tighten.
Team exceeds Enterprise Starter 50-user limit
Enterprise Starter capped at 50 users. Growing team must upgrade to full Enterprise at $59/user - sudden cost jump. No intermediate tier available.
Vendor stability concerns during renewal
Layoffs and product discontinuation raise questions about company future. Enterprise customers face vendor risk assessment. May choose more stable alternatives at renewal.
Large repository exceeds context limitations
Extremely large or complex codebases require iterative prompting for accuracy. Context windows limit AI effectiveness. Not as seamless as marketing suggests for massive monorepos.
AI hallucinations cause bugs in production
Generated code contains subtle bugs from AI hallucinations. Without proper review, bugs reach production. Requires establishing new validation processes for AI-generated code.