# Models and usage — Canopy

> Choose model capability deliberately and interpret token and cost estimates correctly.

The right model depends on the job: a bounded triage task, complex implementation, and independent review ask for different strengths. Compare completed outcomes and rework along with token use. Canopy can summarize usage exposed by supported CLIs, but cost figures are estimates and provider billing remains the authority for actual charges.

Canopy runs installed coding CLIs; their providers control model access, limits, and billing. Check the installed app release for version-specific features. Each link opens the article's clean Markdown version.

## Articles

- [Can Canopy work offline with a local coding agent?](https://canopyide.dev/use-cases/canopy-offline-local-coding-agent-workflow.md): Test local files, Git, services, Preview, search, and a preinstalled CLI against the separate requirement for a local model endpoint.
- [Can Claude Code and Codex agents talk to each other in Canopy?](https://canopyide.dev/use-cases/can-claude-code-and-codex-agents-talk-in-canopy.md): Separate cross-CLI messages from Claude-native agent teams, choose teammate models deliberately, and verify a two-session handoff.
- [Canopy 0.3.4: test tab recovery, OpenCode usage, and Remote](https://canopyide.dev/blog/canopy-034-agent-workflow-release-guide.md): A task-based reading of the v0.3.4 release notes with a short installed-build trial for closed tabs, OpenCode Statistics, and Remote continuity.
- [Canopy vs OpenCode: agent or project workspace?](https://canopyide.dev/use-cases/canopy-vs-opencode-for-coding-agents.md): Compare OpenCode's multi-provider agent, subagents, session stats, and sharing with Canopy's local project workflow around installed CLIs.
- [Coding agent hit a usage limit mid-task? Save the work first](https://canopyide.dev/guides/coding-agent-hit-usage-limit-mid-task.md): Identify the actual limit, preserve the checkout and partial result, then resume, wait, or hand off without repeating work or changing accounts blindly.
- [Did your coding-agent skill actually load? A five-minute test](https://canopyide.dev/guides/check-if-coding-agent-skill-loaded.md): Separate skill discovery, invocation, and useful execution with an explicit test, a natural-language test, and one negative control.
- [How many coding agents should you run at once?](https://canopyide.dev/guides/how-many-coding-agents-to-run-at-once.md): A practical way to size parallel agent work by independent tasks, review capacity, worktrees, local resources, and provider usage.
- [How to evaluate an AI coding workspace on a real project](https://canopyide.dev/guides/evaluate-ai-coding-workspace-on-real-project.md): A repeatable trial for terminal, IDE, and multi-agent workspaces that follows one change from setup through running behavior, review, resume, and accepted result.
- [How to reduce coding-agent token usage without losing the result](https://canopyide.dev/guides/reduce-coding-agent-token-usage.md): Scope tasks, manage growing context, choose models deliberately, and compare accepted work rather than chasing a low token count.
- [How to run multiple coding agents without losing track](https://canopyide.dev/guides/run-multiple-coding-agents.md): A two-agent workflow with task ownership, separate worktrees, service ports, attention checks, and a final review decision.
- [How to set a practical budget for parallel coding agents](https://canopyide.dev/guides/budget-parallel-coding-agents.md): Decide how many agents to launch, assign each an output and stop point, then track plan usage, estimates, review load, and accepted work.
- [How to stop an endless AI code review loop](https://canopyide.dev/guides/stop-endless-ai-code-review-loop.md): When each review pass finds another issue, triage findings against the same acceptance checks, verify the latest commit, and decide what blocks this PR.
- [Keep coding-agent context useful between sessions](https://canopyide.dev/guides/keep-agent-context-between-sessions.md): Make a handoff that preserves the task, decisions, branch, and evidence without carrying an entire chat transcript forever.
- [Measure AI coding cost per accepted change](https://canopyide.dev/guides/measure-ai-coding-cost-per-accepted-change.md): A repeatable worksheet for comparing model choices by accepted work, review time, retries, token categories, and provider charges.
- [Run OpenCode with a local Ollama model in Canopy](https://canopyide.dev/guides/opencode-ollama-local-model-in-canopy.md): Set up a local model route, account for OpenCode V1 and V2 differences, verify context and tool use, then test an offline coding task in Canopy.
- [Share agent work with a team without sharing model accounts](https://canopyide.dev/guides/share-ai-agent-context-without-sharing-model-accounts.md): A practical boundary map for shared project context, CLI profiles, team collaboration, model access, and billing in Canopy.
- [Share project instructions across Claude Code and Codex](https://canopyide.dev/guides/share-instructions-skills-mcp-between-claude-code-and-codex.md): Place durable facts, CLI-specific instructions, reusable skills, and MCP connections where each agent can actually use them.
- [Should I use a cheaper AI model first, then a stronger model to review?](https://canopyide.dev/use-cases/small-model-first-pass-strong-model-review.md): Choose current Claude or OpenAI models for triage, implementation, and review; copy the handoff prompts and compare complete task costs.
- [Switch from Claude Code to Codex without losing the task](https://canopyide.dev/guides/switch-coding-agents-without-losing-context.md): A cross-CLI handoff when usage limits, model choice, or review needs change: preserve the goal, branch, evidence, and next action.
- [Understanding AI coding agent usage and estimated cost](https://canopyide.dev/blog/understanding-ai-coding-agent-costs.md): Read Canopy's CLI, model, session, plan-limit, and estimated-cost views without mistaking a token estimate or stale quota snapshot for a provider charge.
- [When a coding-agent session suddenly burns through usage](https://canopyide.dev/guides/diagnose-runaway-coding-agent-session-usage.md): Pause the session, separate context growth from repeated work and stale estimates, then restart with a smaller verified scope.
- [Which coding-agent CLIs work in Canopy, and what differs?](https://canopyide.dev/guides/coding-agent-cli-support-in-canopy.md): A dated, source-backed comparison of launch, resume, session signals, shared context, and file attribution across seven CLIs.
- [Why cheaper AI tokens can make a coding task more expensive](https://canopyide.dev/blog/cheaper-tokens-more-expensive-coding-task.md): A worked task-cost comparison that counts retries, cache categories, human review, and accepted code instead of ranking models by input-token price.
- [Why did Claude Code forget my task after compaction?](https://canopyide.dev/guides/claude-code-forgot-task-after-compaction.md): Recover a Claude Code task after /compact by checking the actual branch, diff, tests, and lost instructions before giving the agent one verified next step.
- [Why does my coding agent show so many sent tokens?](https://canopyide.dev/guides/why-coding-agent-sent-tokens-are-so-high.md): Understand large input totals, growing conversation context, cache reads and writes, output tokens, and the checks to make before changing models.
- [Your AI coding cost dashboard is not your bill](https://canopyide.dev/blog/your-ai-coding-cost-dashboard-is-not-your-bill.md): Separate the three ledgers behind a coding-agent dashboard: model consumption, plan capacity, and actual provider charges.