--- title: "Compare how six AI coding assistants keep project memory between sessions" description: "See how Copilot, Cursor, Windsurf, Claude Code, Cody, and Tabnine carry project context across sessions, then run one test that shows what survives." canonical: "https://scalewithsearch.com/articles/ai-coding-assistants-memory-comparison" date: "2026-01-28" modified: "2026-09-25" --- ## Site navigation - [Scale With Search](https://scalewithsearch.com/) - Real estate - Real estate - [Real estate](https://scalewithsearch.com/for/real-estate) - Work - Start here - [Send your brief](https://scalewithsearch.com/work#send-your-brief) - [Prepare your six-question brief](https://scalewithsearch.com/work#prepare-your-six-question-brief) - Build - [Site build, content library with SEO, signal desk](https://scalewithsearch.com/work) - For your business - Trades and home services - [Auto body and collision shops](https://scalewithsearch.com/for/auto-body-and-collision-shops) - [Foundation and home repair contractors](https://scalewithsearch.com/for/foundation-and-home-repair) - [Garage door and fencing contractors](https://scalewithsearch.com/for/garage-door-and-fencing-contractors) - [HVAC contractors](https://scalewithsearch.com/for/hvac-contractors) - [Janitorial and commercial cleaning companies](https://scalewithsearch.com/for/janitorial-and-commercial-cleaning) - [Locksmiths](https://scalewithsearch.com/for/locksmiths) - [Moving companies](https://scalewithsearch.com/for/moving-companies) - [Pest control companies](https://scalewithsearch.com/for/pest-control-companies) - [Plumbing and electrical contractors](https://scalewithsearch.com/for/plumbing-and-electrical-contractors) - [Restoration and water or fire damage companies](https://scalewithsearch.com/for/restoration-and-water-fire-damage) - [Roofing companies](https://scalewithsearch.com/for/roofing-companies) - [Towing companies](https://scalewithsearch.com/for/towing-companies) - [Tree services and landscaping companies](https://scalewithsearch.com/for/tree-services-and-landscaping) - [Solar installers](https://scalewithsearch.com/for/solar-installation) - [General contractors](https://scalewithsearch.com/for/general-contractors-and-construction) - [Paving, concrete, and flooring contractors](https://scalewithsearch.com/for/paving) - Practices and professional services - [Bookkeeping and tax practices](https://scalewithsearch.com/for/bookkeeping-and-tax-practices) - [Dental practices](https://scalewithsearch.com/for/dental-practices) - [Family and criminal defense law firms](https://scalewithsearch.com/for/family-and-criminal-defense-law-firms) - [Med spas and aesthetics practices](https://scalewithsearch.com/for/med-spas-and-aesthetics) - [Personal injury law firms](https://scalewithsearch.com/for/personal-injury-law-firms) - [Veterinary clinics](https://scalewithsearch.com/for/veterinary-clinics) - [Gyms and fitness studios](https://scalewithsearch.com/for/fitness) - [Therapy and outpatient health practices](https://scalewithsearch.com/for/therapy-and-outpatient-health) - [Medical billing companies](https://scalewithsearch.com/for/medical-billing) - [Insurance agencies](https://scalewithsearch.com/for/insurance-agencies) - [Financial advisors](https://scalewithsearch.com/for/financial-advisors) - [Property management companies](https://scalewithsearch.com/for/property-management) - [Recruiting and staffing agencies](https://scalewithsearch.com/for/recruiting-and-staffing) - [Architects and interior designers](https://scalewithsearch.com/for/architects-and-interior-designers) - [Logistics and supply chain companies](https://scalewithsearch.com/for/logistics-and-supply-chain) - Agencies, MSPs, and manufacturing - [IT and managed service providers](https://scalewithsearch.com/for/it-and-managed-service-providers) - [Machine shops and precision manufacturers](https://scalewithsearch.com/for/machine-shops-and-precision-manufacturing) - [Marketing agencies and freelancers](https://scalewithsearch.com/for/marketing-agencies-and-freelancers) - [SEO agencies and consultants](https://scalewithsearch.com/for/seo-agencies-and-consultants) - [Small manufacturers and fabricators](https://scalewithsearch.com/for/small-manufacturers-and-fabricators) - Restaurants, shops, studios, and nonprofits - [Restaurants and hospitality businesses](https://scalewithsearch.com/for/restaurants-and-hospitality) - [Retail stores and ecommerce sellers](https://scalewithsearch.com/for/retail-and-ecommerce) - [Photographers, event planners, and travel agents](https://scalewithsearch.com/for/photographers) - [Churches and nonprofits](https://scalewithsearch.com/for/churches-and-nonprofits) - [All industries](https://scalewithsearch.com/for/) - Learn - For your office - [Office job guides](https://scalewithsearch.com/guides/) - [Browser calculators](https://scalewithsearch.com/tools/) - Start here - [How it works](https://scalewithsearch.com/how-it-works) - [Free Starter Kit](https://scalewithsearch.com/kit/business-memory-starter-kit.zip) - [Synthetic specimen](https://scalewithsearch.com/specimen/working-session-specimen.zip) - Guides - [The Complete Guide to Business Memory for AI Agents](https://scalewithsearch.com/articles/business-memory-for-ai-agents-guide) - [The Complete Small-Business Guide to AI Agent Governance](https://scalewithsearch.com/articles/ai-agent-governance-guide-small-business) - [The Complete Guide to Leaving Vendor AI Memory](https://scalewithsearch.com/articles/leaving-vendor-ai-memory-guide) - Articles by cluster - [Business memory](https://scalewithsearch.com/articles/business-memory-for-ai-agents-guide) - [Agent governance](https://scalewithsearch.com/articles/ai-agent-governance-guide-small-business) - [Migration and ownership](https://scalewithsearch.com/articles/leaving-vendor-ai-memory-guide) - For machines - [llms.txt](https://scalewithsearch.com/llms.txt) - [llms-full.txt](https://scalewithsearch.com/llms-full.txt) - [Machine view](https://scalewithsearch.com/?view=machine) - Company - Evidence - [Proof](https://scalewithsearch.com/proof) - Company - [About](https://scalewithsearch.com/about) # Compare how six AI coding assistants keep project memory between sessions. Every AI coding assistant claims to understand your codebase. Analysis of files during a session is one thing. Memory of your project across sessions is another. Most tools focus on code completion. They read your open files, suggest code, and forget the session when you close the IDE. Some tools go further and keep context between sessions. This comparison covers six assistants: GitHub Copilot, Cursor, Windsurf, Claude Code, Cody, and Tabnine. It looks at memory only: what each tool keeps about your project, your conventions, and your preferences from one session to the next. This comparison was first written in January 2026 and checked again on 2026.09.25. Vendor features change monthly, so each claim below carries its date or its source. Check the vendor's current docs before you decide. ## Separate three kinds of memory Vendors use "memory" for three different mechanisms. Keep them apart when you compare tools: 1. **In-session context.** The tool indexes or reads files while you work. It can find a distant file, but the knowledge ends with the session. 2. **Instruction files you write.** A rules or memory file in the repository or your home folder. The tool reads it at the start of each session. You can read, diff, and correct it. 3. **Automatic memory the tool writes.** The tool captures facts from your work and stores them for later sessions. It needs less effort, and it needs a review path, because you did not write it. A tool can offer more than one kind. The table later in this page names which kinds each tool offers. ## GitHub Copilot: instruction files plus repository memory in preview Copilot has long read custom instruction files, such as `.github/copilot-instructions.md` in a repository. Personal instruction files also exist. Before 2026, those files were Copilot's only way to carry context between sessions. GitHub put agentic memory for Copilot into public preview on [January 15, 2026](https://github.blog/changelog/2026-01-15-agentic-memory-for-github-copilot-is-in-public-preview/). The changelog describes this behavior: - Copilot builds a repository-specific memory from key insights about your codebase. - When an agent starts a new session, it retrieves recent memories for the target repository. - Before it applies a memory, the agent checks the cited code locations to verify it. - The feature works in Copilot CLI, the coding agent, and code review. - Repository owners can review and delete stored memories in `Repository Settings > Copilot > Memory`. The memory is repository-scoped. Personal preferences, domain knowledge, and business logic still need an instruction file. The guide on [why Copilot forgets between chats](/articles/why-does-copilot-forget) separates these repository instructions from Microsoft 365 Copilot memory. ## Cursor: project index plus rules Cursor indexes your project for semantic search. When you ask a question, its agent can pull relevant snippets from files you have not opened. Cursor is strong inside a session for that reason. Across sessions, Cursor relies on Rules: persistent, reusable context that it adds at the prompt level. Language models keep no memory between completions, so Rules supply the continuity. Cursor also supports MCP servers, which can connect external memory tools such as [Basic Memory](https://docs.basicmemory.com/integrations/cursor/) and Recallium for persistence across projects. Cursor shipped a native Memories feature in 2025 releases and has changed it since. Confirm its current status in the Cursor changelog. For a direct comparison of rules and repository continuity, see [Claude Code vs Cursor for project memory](/articles/claude-code-vs-cursor-project-memory). ## Windsurf: Cascade agent with rules and memories Windsurf is an agentic IDE built on a VS Code fork. It came from the company formerly named Codeium, and Cognition acquired it in July 2025. Its Cascade agent runs multi-step tasks: it executes terminal commands, creates files, and refactors routes. Tab and Supercomplete give fast autocomplete with awareness of the terminal. Windsurf supports workspace rules files, with one Markdown file per rule in `.windsurf/rules/`. As of 2026.09.25, the Windsurf docs redirect to Cognition's Devin docs, which prefer `.devin/rules/` and keep `.windsurf/rules/` as a fallback. Cascade can also generate memories during a conversation and reuse them in later sessions. The same docs say that memories apply to the legacy Cascade agent only. The newer default agent, Devin Local, does not persist memories. Model choices and plan limits change often, so this page omits them. ## Claude Code: memory files you write and review Claude Code is Anthropic's agentic coding tool. It runs in the terminal and also in IDE extensions, a desktop app, and the web. It reads and edits files and runs commands with built-in tools. The Model Context Protocol (MCP) connects additional external tools. Its memory is the [`CLAUDE.md` system](https://code.claude.com/docs/en/memory): Markdown files that Claude Code reads at the start of each session. The files layer: - `~/.claude/CLAUDE.md` holds your personal rules for every project. - `CLAUDE.md` in the project root holds shared project context and goes into Git. - `CLAUDE.md` in a subdirectory holds rules for that part of the tree and loads when Claude reads files there. - `CLAUDE.local.md` in the project root holds your personal project preferences. Add it to `.gitignore` so it stays out of Git. Claude Code reads files from the working directory upward and combines the layers. More specific files add detail for their part of the tree. A file can import another file with `@path/to/import`. This is explicit, persistent memory. You decide what the assistant keeps, and you can diff every change. Claude Code also keeps auto memory: notes that Claude writes from your corrections and preferences. The notes live per repository in `~/.claude/projects//memory/`, with a `MEMORY.md` index. Each session loads the first 200 lines or 25KB of that index (Claude Code docs, checked 2026.09.25). The notes are plain Markdown, so you can read, edit, and delete them. It is not unlimited. Each file loads into the context window, so a long file costs tokens and attention in every session. It is also not enforcement: Anthropic describes `CLAUDE.md` content as context, not configuration. Keep the files short, and use permissions and hooks for hard limits. [Claude-Mem alternatives for coding agents](/articles/claude-mem-alternatives-coding-agents) scores add-on memory layers against this baseline. ## Cody: codebase search, not session memory Cody by Sourcegraph builds on Sourcegraph's code search. It indexes the whole codebase and uses the code graph to answer questions about relationships between components. It is strong on large codebases and distant files. Cody does not keep cross-session memory. Its context is the codebase, not your project decisions or preferences. Sourcegraph ended Cody Free and Cody Pro on July 23, 2025, and kept Cody Enterprise. A different product, Cody by MeetCody, is a business assistant. This comparison covers Cody by Sourcegraph. ## Tabnine: team patterns in a private deployment Tabnine is the privacy-focused option. It can run fully on-premises, which suits companies with strict data policies. Tabnine describes its Enterprise Context Engine as learning your organization's architecture, frameworks, and coding standards. It also markets zero code retention, SOC 2 compliance, and GDPR alignment. Treat these as vendor claims, and check them against Tabnine's current trust documents. The Context Engine is the closest Tabnine comes to persistent memory. It learns team patterns over time. It does not keep project decisions or personal preferences. Supported models change often, so this page omits them. Check Tabnine's current model list. ## Memory comparison table | Tool | Memory type | Carries context across sessions | How it persists | |---|---|---|---| | GitHub Copilot | Repository insights plus instruction files | Yes; repository memory in preview since January 2026 | Automatic capture with citation checks; instruction files | | Cursor | Project index plus Rules | Through Rules and MCP plugins | Rules files; external memory servers | | Windsurf | Rules plus Cascade memories | Yes | Workspace rules; memories in the legacy Cascade agent only | | Claude Code | Explicit memory files plus auto memory | Yes | `CLAUDE.md` files you write, with imports; notes Claude writes | | Cody | Codebase graph and search | No | Index only | | Tabnine | Enterprise Context Engine | Team patterns only | Learned coding standards | Prices change too often for a comparison page. Each vendor's pricing page is the source. ## Feature comparison table | Tool | Autocomplete | Chat | Agent features | Best fit | |---|---|---|---|---| | GitHub Copilot | Yes | Yes | CLI agent and coding agent | Teams that live in GitHub | | Cursor | Yes | Yes | Multi-file edits | An AI-first IDE | | Windsurf | Yes (Tab and Supercomplete) | Yes | Cascade multi-step tasks | Autonomous coding tasks | | Claude Code | No | Yes | File edits, terminal commands, MCP tools | Explicit, reviewable project memory | | Cody | Yes | Yes | Basic | Search across very large codebases | | Tabnine | Yes | Yes | Basic | Private enterprise deployment | ## Match the tool to the job If you want fast autocomplete and live in GitHub, Copilot fits. Its repository memory, in preview since January 2026, learns patterns over time. If you want an AI-first IDE with strong in-session context, Cursor fits. Add Rules, or an MCP memory server, for continuity. If you want an agent that runs multi-step coding tasks, Windsurf's Cascade fits. It keeps rules, and the legacy Cascade agent keeps memories. If you want to write down exactly what the assistant should know, Claude Code fits. You keep the memory in files, and you can read and diff it. [Prompt caching for large codebases](/articles/ai-coding-assistants-prompt-caching-large-codebases) covers the token cost of loading that context on every run. If you need search across a very large codebase, Cody fits, now through Sourcegraph's enterprise offering. If you need strict privacy, Tabnine fits, with team-level patterns and on-premises deployment. ## Analysis is not memory A tool that analyzes your project finds the answer again each session. A tool with memory starts the session with the answer. The most inspectable memory is a file you write: you can read it, diff it, and fix it. Automatic memory saves effort, but it needs a review path for what the tool wrote on its own. Memory features change often. ChatGPT has expanded its Memory feature several times since launch. Claude's context window grew from 100,000 tokens with Claude 2 to 200,000 with Claude 2.1, both in 2023. From August 2025, Claude Sonnet 4 offered a 1,000,000-token window in beta on the API. A file-based record persists across those changes, because it lives in your repository. ## Run one memory test on any tool Run this test before you trust a tool's memory claim: 1. Write three project rules and one harmless unique fact, such as `CANARY-7Q`, into the tool's instruction file. 2. Start a new session and ask the tool to state the rules and the fact. 3. Change one rule, start another new session, and confirm that the tool follows the new rule. 4. Check what the tool stored on its own, and confirm that you can review and delete it. 5. Remove the canary and confirm that no later session repeats it. The tool passes when it follows the current rules, drops the removed fact, and shows you everything it stored. ## Use more than one tool with one set of rules Many teams use ChatGPT for quick questions and Claude Code for work that needs project or business context. That split works if the rules live in one place. Keep one canonical rules file, and point each tool's instruction file at it or copy it on change. [How to keep context across AI tools](/articles/persistent-context-across-ai-tools) shows the pattern. ## Related: AI memory - [Compare Windsurf and Claude Code on how each keeps project context](/articles/windsurf-vs-claude-code) - [Which AI Coding Assistants Support Prompt Caching for Large Codebases?](/articles/ai-coding-assistants-prompt-caching-large-codebases) - [Claude Code vs Cursor for Project Memory: Rules, Decisions, and Repository Continuity](/articles/claude-code-vs-cursor-project-memory) ---- ```text .|########||. .|########||. .|########||. |##||. .||##|. |##||. .||##|. |##||. .||##|. |#|. .|#|. |#|. .|#|. |#|. .|#|. |#| |#| |#| |#| |#| |#| .#| |#. .#| |#. .#| |#. .#. .#| .#. .#| .#. .#| |#. .#| |#. .#| |#. .#| |# ...... #| |# ...... #| |# ...... #| .# ||#########| #| .# ||#########| #| .# ||#########| #| .#. |######||######|. .#. .#. |######||######|. .#. .#. |######||######|. .#. #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# #. .##|###|##|#|######| .# || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || || |##|#||||||||||||#||#| || # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. # |#||||||||||||||||||||#| #. || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # # .#||||||||||||||||||||||||#| # || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || || ....|||#||||||##|#|||#|#||##||||....|. || #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# #. . ....|# ....#||| ||| .#||| ....#. .# # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. # . ||||#| .#####|| . .#| .#||| ||||# #. .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || .| ||||| #. |#|||||. || #. |#||. .||||| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || || |.... #. ....|#. |. ...|. ....|| || #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# #. .||||||##||||||##||####|||||||#||||||#| .# # .||####|###########################|. # # .||####|###########################|. # # .||####|###########################|. # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#||||||.#.|| # |. ..# #| #|||||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||||| .. |# ## |#| ...#|#|||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#|||#|# .# .#| #| ##|.#..#|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # .#||||||############|######|||||#| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # # |...||#||||||#||||#|#||||||#|||||#||| |.| # #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# #. .. ||||# .|||##|. |#| .|| || .|||#. #| # .# || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || | ...|| ...##| | #| .|. |. ##### .. .| || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || || ||||| || ||||#| . |. |. |#. ||||| |#| || || .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. .# |.....#|....|#.||||.|||##.|#|....||.#||.#. #. #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# #.|||||||#############################| ||| .# || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || || ##||#||#||||||#||#||#||#||##. || # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # # .#|||||||||#||#|||||||||||#| # || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || || |#||||||||||||||||||||||#| || .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. .# |#||||||||||||||||||||#| #. || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || || |#||||||||||||||||||#| || #. .######|#|#########| .# #. .######|#|#########| .# #. .######|#|#########| .# .#. |#####||||#####| .#. .#. |#####||||#####| .#. .#. |#####||||#####| .#. |# ||########|| #. |# ||########|| #. |# ||########|| #. |# ...... #| |# ...... #| |# ...... #| |#. .#| |#. .#| |#. .#| |#. .#. |#. .#. |#. .#. .#| |#. .#| |#. .#| |#. |#| |#| |#| |#| |#| |#| .|#|. .|#|. .|#|. .|#|. .|#|. .|#|. |##||. .||##| |##||. .||##| |##||. .||##| .||########||. .||########||. .||########||. Scale With Search 2026 [scalewithsearch.com](https://scalewithsearch.com) ```