Glossary
43 terms used across Codexa, each linked to the page that explains it in depth.
A
- Activity history
- A per-project log of notable events (file saves, snapshots, AI runs, deployments) with the actor and timestamp. Shown in the project's Activity tab. Learn more
- Agent
- The Codexa AI agent: a bounded, tool-using loop that plans, reads and edits project files, runs checks in the sandbox and repairs errors. It never runs code inside the web server. Learn more
- Ask mode
- An AI mode that answers questions about your project without editing any files. Useful for understanding code or planning a change before committing to it. Learn more
B
- Build mode
- The default AI mode. The agent plans, creates and edits files, then validates the result with type-check, build or tests in the sandbox. Learn more
C
- Checkpoint
- Another name for a snapshot: a named, immutable copy of every file in a project that can be diffed against the current state and restored in one click. Learn more
- CNAME record
- A DNS record that points your domain (for example app.example.com) at a Codexa deployment host. Custom domains are set up with a CNAME; automated verification is on the roadmap. Learn more
- Codexa hosting
- The built-in static hosting provider. A deployment builds the project in a sandbox and serves the output directory at a Codexa subdomain, with history and rollback. Learn more
- Confirmation required
- The confirmation_required error returned by the SQL runner when a statement would modify or drop data. Re-send the query with confirmDestructive: true to run it. Learn more
- Connection
- A workspace-level integration record (PostgreSQL, Supabase, OpenAI, Anthropic, Stripe, Razorpay, Resend, Vercel). Secrets are encrypted at rest and never returned to the browser. Learn more
- Conversation
- A thread of messages between you and the AI agent inside a project. Each generation run belongs to a conversation. Learn more
- Custom domain
- Your own domain served in front of a Codexa deployment. Plans include a number of custom domains; configure them with a CNAME record. Learn more
D
- Deployment
- A build of a project at a point in time, published to Codexa hosting or Vercel. Deployments have a status (queued, building, deploying, live, failed, rolled_back), a build log and a URL. Learn more
- Diagnostics
- Type and syntax problems reported by the editor for the open file, shown inline and in the problems list. Learn more
- Diff view
- A side-by-side or inline comparison of two versions of a file, used to review snapshot differences and the changes made by an AI run before keeping or reverting them. Learn more
E
- Environment variable
- A named value (often a secret) made available to a project's preview and deployment. Values are encrypted with the platform key before storage and masked in the UI by default. Learn more
F
- Fix mode
- An AI mode focused on a reported error: the agent reads the failing output, locates the cause and applies a targeted repair. Learn more
- Framework
- The project type that decides how it is installed, run and built: vite-react, static or node-express today. Next.js projects are on the roadmap. Learn more
G
- Generation run
- One execution of the AI agent for a prompt. A run records its stage, plan, changed files, token usage, repair attempts and the snapshot taken before it started. Learn more
I
- Idempotent webhook
- A webhook handler that produces the same result no matter how many times the same event is delivered. Codexa records every billing event id in webhook_events and skips duplicates. Learn more
- Import from GitHub
- Creating a project from an existing repository and branch through the GitHub integration. Files are copied into Codexa's database; the repository is remembered for later pushes. Learn more
L
- Live preview
- A running instance of your project inside a sandbox, proxied to a per-session preview URL and shown in the workspace with viewport controls. Learn more
M
- Monaco
- The open-source editor component that powers the Codexa code editor (the same core as Visual Studio Code). Codexa serves it from its own origin rather than a CDN. Learn more
P
- Personal workspace
- The workspace created automatically for every account. It has a single member (you) and holds projects that are not shared with a team. Learn more
- Plan
- A pricing tier (Free, Pro, Team, Enterprise) that sets a workspace's usage limits. Limits are enforced on the server, not only displayed. Learn more
- Presence
- Realtime information about who is viewing which file in a project, delivered over WebSockets. Presence is shipped; concurrent editing is not. Learn more
- Preview proxy
- The host-based proxy in the runner service that routes a preview URL to the right sandbox port and attaches logs. Learn more
- Project
- The unit of work in Codexa: a set of files, settings, environment variables, snapshots, conversations, runtime sessions and deployments inside a workspace. Learn more
Q
- Quota
- The portion of a plan limit you have used this billing period (AI requests and tokens, runtime minutes, deployments, storage, projects, members). Exceeding it returns quota_exceeded. Learn more
R
- Rate limit
- A cap on how many requests an IP or account can make to an endpoint in a time window. Exceeding it returns HTTP 429 with code rate_limited and a Retry-After header. Learn more
- Repair loop
- After the agent edits files it runs a validation command; if that fails it reads the error output and tries a targeted fix, at most twice, before reporting the remaining error to you. Learn more
- Revert
- Restoring a project to the snapshot that was taken before an AI run, discarding every change the run made. Learn more
- Role
- A member's permission level in a workspace: owner, admin, editor or viewer. Roles are checked on the server for every request. Learn more
- Rollback
- Making a previous successful deployment live again without rebuilding. Learn more
- Runner
- The separate service that executes project code in sandboxes, exposes allowlisted commands (install, build, test, typecheck, lint), streams logs and serves previews and static deployments. Learn more
- Runtime session
- A single running sandbox for a project, with a status (starting, running, stopped, error), a preview URL and a heartbeat. Runtime minutes count against your plan. Learn more
S
- Sandbox
- An isolated environment for running project code: an unprivileged user, Linux namespaces, cgroup memory and process limits, and timeouts. Docker and E2B drivers are also available. Learn more
- Session
- A signed-in browser session, stored as an httpOnly cookie whose token is hashed before it is saved. Sessions are what authenticate API requests today. Learn more
- Snapshot
- A named, immutable copy of every file in a project. Codexa takes one automatically before each AI run and before deployments; you can create them manually at any time. Learn more
- SSE (server-sent events)
- The streaming format used by the generation endpoint and runtime logs: a long-lived HTTP response that emits named events as they happen. Learn more
T
- Template
- A starter project with real files and documented dependencies (React + Vite, static site, Node + Express). Every new project begins from a template or a GitHub import. Learn more
- Token (AI)
- The unit AI providers bill by. Codexa records input and output tokens for every run and enforces the plan's monthly token limit. Learn more
W
- Workspace
- The container for projects, members, integrations, usage and billing. Every user has a personal workspace and may belong to team workspaces. Learn more