# CoderFriend AI > A multi-provider AI coding workspace for Visual Studio Code that combines chat, repository context, guarded local and SSH execution, model routing, MCP tools, worktrees, and delegated agents. The public catalog documents 96 feature systems and 449 concrete capabilities. The HTML page is server-rendered, searchable, and uses stable anchors for direct citations. ## Primary resources - [Feature catalog](https://coderfriendai.com/features/): Human-readable catalog with every capability and its complete explanation. - [Complete capability reference](https://coderfriendai.com/features.md): Markdown version of all systems and capabilities for search and AI retrieval. - [Provider and model catalog](https://coderfriendai.com/models/): Supported providers, connection methods, and public model references. - [Help and setup](https://coderfriendai.com/help/): Current app help, provider setup, tools, and usage guidance. - [Complete public knowledge](https://coderfriendai.com/knowledge.md): Published Features, Models, Help, and commerce documentation used by ChatOne. - [Structured retrieval corpus](https://coderfriendai.com/rag.json): Stable document and chunk IDs, source URLs, and current public text. - [Full text compatibility feed](https://coderfriendai.com/llms-full.txt): Alias of the complete public knowledge reference. - [Security policy](https://github.com/africz/coderfriend-ai/blob/main/SECURITY.md): Security reporting and product safety information. - [XML sitemap](https://coderfriendai.com/seo-sitemap.xml): Canonical indexable website URLs. ## Capability groups - **AI workspace:** 19 feature systems and 96 capabilities. - **Living context:** 18 feature systems and 75 capabilities. - **Safe execution:** 32 feature systems and 157 capabilities. - **Models & orchestration:** 27 feature systems and 121 capabilities. ## Feature systems - [Chat, Plan & Agent modes](https://coderfriendai.com/features/#feature-system-1-chat-plan-agent-modes): Move from a quick answer to an implementation plan or an autonomous coding run without changing workspaces. - [Integrated composer cockpit](https://coderfriendai.com/features/#feature-system-2-integrated-composer-cockpit): Shape every request from one compact control surface instead of bouncing between settings screens. - [Live progress & activity](https://coderfriendai.com/features/#feature-system-3-live-progress-activity): See what the Agent is doing as it searches, reads, edits, and verifies the task. - [Queue, steer, delete & stop](https://coderfriendai.com/features/#feature-system-4-queue-steer-delete-stop): Stay in control while an Agent run is active, even when your priorities change mid-task. - [Inline Chat](https://coderfriendai.com/features/#feature-system-5-inline-chat): Ask for a targeted change directly beside the code and review the proposed result in place. - [Inline completion](https://coderfriendai.com/features/#feature-system-6-inline-completion): Complete code with both sides of the cursor, recent edits, repository guidance, and bounded editor context. - [One-click editor actions](https://coderfriendai.com/features/#feature-system-7-one-click-editor-actions): Turn common code tasks into focused prompts from the lightbulb or editor context menu. - [Actionable code blocks](https://coderfriendai.com/features/#feature-system-8-actionable-code-blocks): Use generated code immediately from the conversation instead of manually moving it between surfaces. - [Native VS Code entry points](https://coderfriendai.com/features/#feature-system-9-native-vs-code-entry-points): Reach CoderFriend from the sidebar, command palette, slash menu, or VS Code Chat. - [A workspace that fits you](https://coderfriendai.com/features/#feature-system-10-a-workspace-that-fits-you): Tune the density and typography of the interface without giving up access to advanced controls. - [25-command slash palette](https://coderfriendai.com/features/#feature-system-11-25-command-slash-palette): Control coding actions, modes, models, safety, history, usage, skills, and MCP from a discoverable keyboard-first menu. - [Model Debug workspace](https://coderfriendai.com/features/#feature-system-12-model-debug-workspace): Inspect the bounded request and response trace when a provider, tool call, or streamed result needs a closer look. - [Generated visual artifacts](https://coderfriendai.com/features/#feature-system-13-generated-visual-artifacts): Create raster assets inside the repository and open visual deliverables directly in VS Code for immediate inspection. - [Resumable bounded requests](https://coderfriendai.com/features/#feature-system-14-resumable-bounded-requests): Continue useful work after a deliberate safety limit without replaying completed steps or losing the accumulated task state. - [Searchable in-app Help](https://coderfriendai.com/features/#feature-system-15-searchable-in-app-help): Keep the complete operating guide inside CoderFriend so setup, safety, tools, and advanced workflows are searchable while you work. - [Smart conversation organization](https://coderfriendai.com/features/#feature-system-16-smart-conversation-organization): Turn long-running threads into durable project records with meaningful titles, retained usage, and portable exports. - [Terminal troubleshooting](https://coderfriendai.com/features/#feature-system-17-terminal-troubleshooting): Bring a captured terminal command into a focused explanation or repair workflow. - [Paused debugger explanations](https://coderfriendai.com/features/#feature-system-18-paused-debugger-explanations): Carry an actual paused stack and bounded variables into an explanation request. - [Jupyter notebook cells](https://coderfriendai.com/features/#feature-system-19-jupyter-notebook-cells): Read and edit notebooks as cells while keeping changes reviewable. - [Files, folders & image context](https://coderfriendai.com/features/#feature-system-20-files-folders-image-context): Give a request the exact source material it needs, from one selection to an entire project area. - [Ranked local codebase search](https://coderfriendai.com/features/#feature-system-21-ranked-local-codebase-search): Find relevant project code locally with bounded retrieval designed for useful signal, not indiscriminate context. - [Repository instructions](https://coderfriendai.com/features/#feature-system-22-repository-instructions): Carry project conventions into every answer with instruction files discovered from the repository tree. - [Portable skills](https://coderfriendai.com/features/#feature-system-23-portable-skills): Reuse durable workflows and domain guidance through SKILL.md registries that travel with your setup. - [Current web search](https://coderfriendai.com/features/#feature-system-24-current-web-search): Bring current public information into a request when repository context alone is not enough. - [Local documentation library](https://coderfriendai.com/features/#feature-system-25-local-documentation-library): Index the documentation you trust and call it into a coding conversation with @docs. - [MCP servers & tools](https://coderfriendai.com/features/#feature-system-26-mcp-servers-tools): Connect approved external tools and data sources through Model Context Protocol servers. - [Adaptive context compaction](https://coderfriendai.com/features/#feature-system-27-adaptive-context-compaction): Keep long conversations useful by preserving what matters as the context window fills. - [Conversation history](https://coderfriendai.com/features/#feature-system-28-conversation-history): Treat conversations as durable project work instead of disposable chat tabs. - [Project backup & restore](https://coderfriendai.com/features/#feature-system-29-project-backup-restore): Package important conversations with their supporting material so work can be archived or moved safely. - [Automatic codebase context](https://coderfriendai.com/features/#feature-system-30-automatic-codebase-context): Supply relevant repository snippets automatically when a request needs more than the open editor, without uploading an external vector index. - [Skill authoring & cross-provider invocation](https://coderfriendai.com/features/#feature-system-31-skill-authoring-cross-provider-invocation): Use one task-specific skill system across model providers while preserving each ecosystem’s existing project and user registries. - [Safe public page fetching](https://coderfriendai.com/features/#feature-system-32-safe-public-page-fetching): Retrieve the readable content behind an explicit public URL with bounded redirects, output limits, and credential-aware safeguards. - [Bounded documentation crawler](https://coderfriendai.com/features/#feature-system-33-bounded-documentation-crawler): Turn official documentation sites into a local, searchable reference library with explicit depth, page, and token limits. - [Secret-aware retrieval](https://coderfriendai.com/features/#feature-system-34-secret-aware-retrieval): Search useful project context without silently sweeping credentials, private keys, or likely secret files into model-visible results. - [Durable conversation memory](https://coderfriendai.com/features/#feature-system-35-durable-conversation-memory): Reopen project conversations with their messages, debug trace, attachments, title, and measured usage ready for continued work. - [Approved project memory](https://coderfriendai.com/features/#feature-system-36-approved-project-memory): Keep useful project facts locally and decide which remembered context remains active. - [Shared tools for VS Code AI](https://coderfriendai.com/features/#feature-system-37-shared-tools-for-vs-code-ai): Make CoderFriend workspace-reading tools available to compatible VS Code Language Model API callers. - [Approval profiles](https://coderfriendai.com/features/#feature-system-38-approval-profiles): Choose how much authority the Agent has for each workspace and each kind of task. - [Workspace sandbox boundaries](https://coderfriendai.com/features/#feature-system-39-workspace-sandbox-boundaries): Match execution freedom to the risk of the task with explicit filesystem and command boundaries. - [Atomic file editing](https://coderfriendai.com/features/#feature-system-40-atomic-file-editing): Apply focused changes with tools designed to protect surrounding code and reduce accidental rewrites. - [Checkpoints, diffs & undo](https://coderfriendai.com/features/#feature-system-41-checkpoints-diffs-undo): Review the real workspace impact and recover from an unwanted change without guesswork. - [Semantic code intelligence](https://coderfriendai.com/features/#feature-system-42-semantic-code-intelligence): Use the language server to understand symbols and surface compiler-quality feedback during a task. - [High-signal code review](https://coderfriendai.com/features/#feature-system-43-high-signal-code-review): Review exactly the scope that matters, from a selection to a GitHub pull request. - [Review-to-repair workflow](https://coderfriendai.com/features/#feature-system-44-review-to-repair-workflow): Turn a finding into an isolated repair task and verify the fix before it reaches your main worktree. - [Runtime verification](https://coderfriendai.com/features/#feature-system-45-runtime-verification): Let the Agent prove that an interface works, not merely that the code looks plausible. - [Live Server Agent](https://coderfriendai.com/features/#feature-system-46-live-server-agent): Diagnose and repair remote environments over OpenSSH with explicit roots and rollback-aware tools. - [Request budgets & recovery](https://coderfriendai.com/features/#feature-system-47-request-budgets-recovery): Bound autonomous work by the resources you choose, then continue cleanly after an interruption. - [Agent evaluations & scorecards](https://coderfriendai.com/features/#feature-system-48-agent-evaluations-scorecards): Compare how provider and model combinations perform against a repeatable coding-agent benchmark. - [Exact approval memory](https://coderfriendai.com/features/#feature-system-49-exact-approval-memory): Turn repeated safe actions into explicit reusable rules without granting a model vague command-prefix authority. - [Destructive-command intelligence](https://coderfriendai.com/features/#feature-system-50-destructive-command-intelligence): Classify dangerous shell behavior before execution so approval prompts explain the actual irreversible risk in a complete command line. - [Process-isolated command execution](https://coderfriendai.com/features/#feature-system-51-process-isolated-command-execution): Run project commands inside an explicit filesystem and network boundary with bounded output, time, cancellation, and complete process-tree cleanup. - [Docker-aware host verification](https://coderfriendai.com/features/#feature-system-52-docker-aware-host-verification): Handle Docker-backed repository checks as a distinct host boundary instead of pretending daemon access fits inside an ordinary workspace sandbox. - [Atomic workspace-wide replacement](https://coderfriendai.com/features/#feature-system-53-atomic-workspace-wide-replacement): Apply one exact mechanical wording, label, or import-path change across an explicit file set as a single recoverable transaction. - [Shell-edit checkpoint recovery](https://coderfriendai.com/features/#feature-system-54-shell-edit-checkpoint-recovery): Capture recoverable before-states for files explicitly named by recognized shell-edit commands, even during maximum-trust work. - [Concurrent edit protection](https://coderfriendai.com/features/#feature-system-55-concurrent-edit-protection): Prevent foreground, background, and delegated Agents from silently writing the same path at the same time. - [Diagnostics-guided repair](https://coderfriendai.com/features/#feature-system-56-diagnostics-guided-repair): Feed compiler and language-service evidence back into the Agent loop so edits can be corrected before completion is claimed. - [Evidence-backed completion](https://coderfriendai.com/features/#feature-system-57-evidence-backed-completion): Require the Agent to connect completion claims to edits, diagnostics, tests, runtime observations, or an explicit limitation. - [Isolated browser interaction](https://coderfriendai.com/features/#feature-system-58-isolated-browser-interaction): Test local interfaces and explicitly requested public pages in a fresh browser profile that never borrows personal cookies or credentials. - [SSH connection identity & trust](https://coderfriendai.com/features/#feature-system-59-ssh-connection-identity-trust): Connect to a deliberate OpenSSH target while keeping environment, host, root, capability, and temporary trust visible throughout the session. - [SSH Safe Tools sandbox](https://coderfriendai.com/features/#feature-system-60-ssh-safe-tools-sandbox): Operate on a remote deployment through structured, root-bounded tools while unrestricted shell execution remains unavailable. - [SSH Full Shell authority](https://coderfriendai.com/features/#feature-system-61-ssh-full-shell-authority): Enable unrestricted remote commands only through a separate, explicit capability decision that clearly names its reach beyond the structured file root. - [SSH large-file inspection](https://coderfriendai.com/features/#feature-system-62-ssh-large-file-inspection): Inspect large or rotating remote logs in place instead of downloading an entire multi-megabyte file into the Agent context. - [SSH operational diagnostics](https://coderfriendai.com/features/#feature-system-63-ssh-operational-diagnostics): Investigate a live environment through fixed, inspectable operational tools before considering broader remote shell access. - [SSH managed process lifecycle](https://coderfriendai.com/features/#feature-system-64-ssh-managed-process-lifecycle): Start temporary remote servers across tool calls while preserving request ownership, bounded lifetime, observable output, and verified cleanup. - [SSH atomic backup & recovery](https://coderfriendai.com/features/#feature-system-65-ssh-atomic-backup-recovery): Make remote production changes with per-file recovery points, transactional rollback, and an explicit restore path. - [Resilient Agent execution](https://coderfriendai.com/features/#feature-system-66-resilient-agent-execution): Recover from common model, command, and tool misunderstandings without hiding errors or spending unlimited retries. - [Organization policies & roles](https://coderfriendai.com/features/#feature-system-67-organization-policies-roles): Apply machine-level guardrails that narrow the authority already allowed by the user. - [Project checks before tools](https://coderfriendai.com/features/#feature-system-68-project-checks-before-tools): Run explicitly trusted repository checks before supported Agent tool calls. - [Local audit log & verification](https://coderfriendai.com/features/#feature-system-69-local-audit-log-verification): Keep an optional, inspectable record of tool, approval, command, and hook activity on your machine. - [12 model sources](https://coderfriendai.com/features/#feature-system-70-12-model-sources): Choose local, subscription, or API-backed intelligence from one consistent coding workspace. - [Cross-provider continuity](https://coderfriendai.com/features/#feature-system-71-cross-provider-continuity): Keep the same conversation and project context when a different model is better for the next step. - [Codex with ChatGPT](https://coderfriendai.com/features/#feature-system-72-codex-with-chatgpt): Use the shared Codex login flow for access backed by your ChatGPT account. - [Claude subscription support](https://coderfriendai.com/features/#feature-system-73-claude-subscription-support): Connect Claude Pro or Max through Claude Code as an alternative to API-key billing. - [Secure API connections](https://coderfriendai.com/features/#feature-system-74-secure-api-connections): Connect provider APIs without placing secrets in ordinary workspace settings or source files. - [Portable provider setup](https://coderfriendai.com/features/#feature-system-75-portable-provider-setup): Move provider configuration between environments without rebuilding every preference by hand. - [Model favorites & aliases](https://coderfriendai.com/features/#feature-system-76-model-favorites-aliases): Make a large model catalog feel personal, fast, and task-specific. - [Profiles & guardrails](https://coderfriendai.com/features/#feature-system-77-profiles-guardrails): Save a working style once, then apply it consistently across tasks and providers. - [Prompt caching](https://coderfriendai.com/features/#feature-system-78-prompt-caching): Reuse eligible prompt context with supported providers to reduce repeated processing. - [Usage & cost visibility](https://coderfriendai.com/features/#feature-system-79-usage-cost-visibility): Understand what each conversation consumes without leaving the coding flow. - [Provider accounts & spending budgets](https://coderfriendai.com/features/#feature-system-80-provider-accounts-spending-budgets): Pair usage data with account-level limits so cost remains an explicit engineering choice. - [Managed Git worktrees](https://coderfriendai.com/features/#feature-system-81-managed-git-worktrees): Give concurrent or risky work an isolated branch and filesystem without leaving CoderFriend. - [Background Agent runs](https://coderfriendai.com/features/#feature-system-82-background-agent-runs): Let longer tasks continue outside the foreground conversation while you keep working. - [Bounded specialist subagents](https://coderfriendai.com/features/#feature-system-83-bounded-specialist-subagents): Delegate focused investigation without handing every helper unrestricted control of the workspace. - [Shared plans & delegated review](https://coderfriendai.com/features/#feature-system-84-shared-plans-delegated-review): Coordinate larger tasks through visible plans, reusable artifacts, and deliberate specialist handoffs. - [Provider-attributed commit messages](https://coderfriendai.com/features/#feature-system-85-provider-attributed-commit-messages): Generate a concise commit message from the intended diff with clear model attribution. - [Live provider model catalogs](https://coderfriendai.com/features/#feature-system-86-live-provider-model-catalogs): Build each picker from the models the selected provider actually exposes, while keeping provider catalogs isolated from one another. - [Provider-specific model memory](https://coderfriendai.com/features/#feature-system-87-provider-specific-model-memory): Remember the right model and generation controls for each provider and working mode instead of forcing one global compromise. - [Provider authentication recovery](https://coderfriendai.com/features/#feature-system-88-provider-authentication-recovery): Recognize expired or missing provider authentication and guide the user back to a valid connection without losing the active workflow. - [Durable usage accounting](https://coderfriendai.com/features/#feature-system-89-durable-usage-accounting): Account for model work at the provider boundary and preserve measured totals with the conversation that produced them. - [Live pricing & account budgets](https://coderfriendai.com/features/#feature-system-90-live-pricing-account-budgets): Combine an editable exact-model rate registry with provider account links, measured spend, and configurable budget periods. - [Dependency-aware task engine](https://coderfriendai.com/features/#feature-system-91-dependency-aware-task-engine): Represent larger work as a shared task graph so ready work, blocked dependencies, artifacts, and completion state remain inspectable. - [Live subagent console](https://coderfriendai.com/features/#feature-system-92-live-subagent-console): See delegated specialists as active parts of the request instead of receiving an unexplained combined answer at the end. - [Hierarchical delegation budgets](https://coderfriendai.com/features/#feature-system-93-hierarchical-delegation-budgets): Bound each specialist and the parent-wide pool so parallel investigation cannot silently exceed the request’s intended cost or time. - [LM Studio local models](https://coderfriendai.com/features/#feature-system-94-lm-studio-local-models): Use models served by LM Studio on your own machine in the same CoderFriend workspace. - [Agent work beyond the editor window](https://coderfriendai.com/features/#feature-system-95-agent-work-beyond-the-editor-window): Optionally continue compatible background work in a detached process and review its recorded outcome later. - [GitHub issue-to-review workflow](https://coderfriendai.com/features/#feature-system-96-github-issue-to-review-workflow): Connect issue context, local branch work, pull-request feedback, and explicit publishing actions through the GitHub CLI.