Long conversations and context compaction
How long threads work with Fast and Think modes. Model brands change; product modes and plan access are the durable truth.
Long compliance threads can grow large. ISMS Copilot may compact older turns so the assistant keeps working without failing on context size. Compaction is a product reliability feature. It is not tied to a single vendor model name forever.
Use the mode guide, not brand names
For which plan can use Fast, Think, and Beyond, and how providers route today, read:
Do not treat older Ferndesk text that said "Think = Claude Opus only" or "Fast = Claude Sonnet only" as current. Paid default routing (mid-2026, ADP off) can be xAI Grok with Anthropic backup; ADP uses EU Mistral. Free/Essential use economy routing.
Practical tips
- Prefer one task per conversation when you are on Free (credit and message caps are tight).
- On Think / Beyond, longer multi-step work is expected; still start a new thread when you switch clients or frameworks.
- Upload core documents early in a thread when the whole analysis depends on them.
- If you hit a session usage limit, wait for the next 4-hour UTC bin or upgrade / trial. See What to do when you hit your usage limit.
Plan access (summary)
| Plan | Think | Beyond / web research |
|---|---|---|
| Free | No | No |
| Essential (grandfathered) | Yes | No (default package) |
| Plus+ | Yes | Yes |