autogpt-platform-beta-v0.6.55

# 🚀 Release `autogpt-platform-beta-v0.6.55`

**Date:** april 2026

---

## 🔥 What's New?

### New Features
- **#12720** - Add Graphiti temporal knowledge graph memory for CoPilot (by @ntindle)
- **#12711** - Hide session-level dry_run from LLM in CoPilot (by @majdyz)
- **#12732** - Support multiple questions in ask_question tool for CoPilot (by @majdyz)
- **#12726** - Per-model cost breakdown, cache token tracking, and OrchestratorBlock cost fix in platform admin (by @majdyz)
- **#12747** - Upgrade claude-agent-sdk to 0.1.58 with OpenRouter compatibility and cost controls (by @majdyz)
- **#12787** - Add for_agent_generation flag to find_block in CoPilot (by @majdyz)
- **#12786** - Standard/advanced model toggle with Opus rate-limit multiplier (by @majdyz)
- **#12757** - Enhance cost dashboard with token breakdown and averages in platform admin (by @majdyz)

### Enhancements
- **#12725** - Enable LLM prompt caching and harden user_context injection in CoPilot (by @majdyz)
- **#12758** - Enable cross-user prompt caching via SystemPromptPreset (by @majdyz)
- **#12790** - Make system prompt fully static for cross-user prompt caching (by @majdyz)

### UI/UX Improvements
- **#12793** - Align ModelToggleButton styling and add execution ID filter to platform cost page (by @majdyz)
- **#12792** - Load older chat messages reliably and preserve scrollback across turns (by @kcze)
- **#12770** - Tighten artifact preview behavior (by @ntindle)

### Bug Fixes
- **#12636** - P0 guardrails, transient retry, and security hardening in CoPilot (by @majdyz)
- **#12718** - Unify CoPilot auth headers and propagate impersonation header (by @majdyz)
- **#12759** - Deduplicate SSE-replayed messages by content fingerprint (by @majdyz)
- **#12763** - Strip `<internal_reasoning>` tags from Sonnet response stream (by @majdyz)
- **#12739** - Keep credential setup inline on run and schedule paths (by @majdyz)
- **#12768** - Remove extra blank line in platform_cost_test.py (by @Bentlybro)
- **#12777** - Fix T2+ --resume by using CLI native session file (by @majdyz)
- **#12774** - Fix sub-agent session hang and orphan on E2B API stall (by @majdyz)
- **#12761** - Store mode per session so indicator updates on switch (by @majdyz)
- **#12750** - Unified MCP file tools (Read/Write/Edit) to prevent truncation data loss (by @majdyz)
- **#12762** - Baseline cost tracking and cache token display (by @majdyz)
- **#12782** - Robust context fallback — upload gate, gap-fill, and token-budget compression (by @majdyz)
- **#12735** - Charge per LLM iteration and per tool call in OrchestratorBlock (by @majdyz)
- **#12789** - Null-safe token accumulation for OpenRouter null cache fields (by @majdyz)
- **#12766** - Abort SSE stream and disconnect backend listeners on session switch (by @0ubbe)
- **#12788** - Idempotency guard and frontend dedup fix for duplicate messages (by @majdyz)
- **#12795** - Set session_id on mode-switch T1 to enable --resume on subsequent SDK turns (by @majdyz)

### Documentation
- **#12695** - Add coverage guidance to pr-address skill (by @Otto-AGPT)

### Dependencies & Maintenance
- **#12682** - Add Playwright PR smoke suite with seeded QA accounts (by @Abhi1992002)
- **#12778** - Untrack auto-generated API client model files (by @majdyz)

---

## 🎉 Thanks to Our Contributors!

As well as our returning contributors:
- @majdyz
- @0ubbe
- @kcze
- @ntindle
- @Bentlybro
- @Abhi1992002
- @Otto-AGPT

---

## 📥 How to Get This Update

To update to this version, run:

```bash
git pull origin autogpt-platform-beta-v0.6.55
```

Or download it directly from the  
[Releases page](https://github.com/Significant-Gravitas/AutoGPT/releases).

For a complete list of changes, see the [Full Changelog](https://github.com/Significant-Gravitas/AutoGPT/compare/autogpt-platform-beta-v0.6.54...autogpt-platform-beta-v0.6.55).

---

## 📝 **Feedback and Issues**
If you encounter any issues or have suggestions, please [join our Discord](https://discord.gg/autogpt) and let us know!