Eigent Release Notes v0.0.86: Triggers, Local Models & a More Polished Platform

This sprint delivers one of the most requested features — automated triggers — alongside expanded local model support, smarter skill handling, and a wave of UI polish. These updates make Eigent more automatable, more accessible to local-first users, and more refined across the board. This release continues to reinforce Eigent's direction as an open, extensible multi-agent cowork platform.

As always, these improvements are community-driven. Let's dive in.

---

⏰ Schedule & Webhook Triggers

This is a game-changer for automation.

Huge thanks to @a7m-1st and @Douglasymlai for shipping the full Triggers system end-to-end.

Eigent now supports scheduled triggers and webhook triggers, letting agents run tasks automatically on a schedule or in response to external events — without manual intervention.

What's new:

• Create schedule-based triggers with cron expressions
• Create webhook triggers with auto-generated URLs
• Webhook dialog shown after creation for easy integration
• Live execution logs panel (slide-in layout, not overlay)
• Toggle button to open/fold execution logs with Framer Motion animation
• Execution state persisted via localStorage
• Default date set to now with UTC conversion before sending to server
• Trigger server migration updates for backend stability

This makes Eigent a true automation-capable agent platform, not just a chat interface.

🔗 PR: https://github.com/eigent-ai/eigent/pull/823 🔗 PR: https://github.com/eigent-ai/eigent/pull/1412 🔗 PR: https://github.com/eigent-ai/eigent/pull/1415 🔗 PR: https://github.com/eigent-ai/eigent/pull/1416

---

🦙 Native LLaMA.cpp Local Provider Support

Local-first AI just got easier.

Shoutout to @it-education-md for adding native LLaMA.cpp as a dedicated Local Model provider in Eigent.

Eigent now integrates directly with llama-server, bringing true local inference without any cloud dependency.

What's new:

• LLaMA.cpp provider entry in Settings → Local Models
• llama-server integration with /v1/health and /v1/models endpoints
• Backend platform compatibility mapping (llama.cpp → openai-compatible-model)
• Full selection, default, save, and reset flow
• Model suggestion updates with tests
• Documentation updates for Local Model / Quick Start

For users who want full data privacy or work in air-gapped environments, this is a major step forward.

🔗 PR: https://github.com/eigent-ai/eigent/pull/1346

---

🔍 Search Toolkit Now Available Across All Agents

Search is no longer browser-only.

Thanks to @bytecii, the Search toolkit is now enabled for Developer, Document, Multi-Modal, and Social Media agents — not just the Browser agent.

What's new:

• Search toolkit wired into all non-browser worker agents
• Fixed a bug where SearchToolkit was imported but never connected in multi-modal agent
• Agent name attribution added to toolkit activation events
• Agent name displayed in the toolkit UI

This means every agent can now search the web independently, making multi-agent workflows significantly more capable.

🔗 PR: https://github.com/eigent-ai/eigent/pull/1364

---

🎵 Memory-Efficient Streaming Multimedia Playback

Performance upgrade for media-heavy workflows.

Shoutout to @Zephyroam for replacing data URLs with file:// URLs for audio and video playback, dramatically reducing memory usage.

What's improved:

• Streaming playback via file:// URLs instead of loading entire files into memory as data URLs
• Significantly lower memory footprint for large media files
• Smoother playback experience overall

This is especially impactful when agents generate or process large audio/video outputs.

🔗 PR: https://github.com/eigent-ai/eigent/pull/1401

---

🧩 Smarter Skill Access Control

Skills now work reliably across all agents.

Thanks to @fengju0213 for refactoring skill access control to use CAMEL's allowed_skills filtering and fixing a critical bug where single-agent scoped skills couldn't be accessed.

What's fixed:

• Frontend now stores agent scope using stable agent IDs instead of display labels
• Backend computes allowed_skills from skills-config.json and passes to CAMEL SkillToolkit
• Removed fragile Eigent-specific post-scan filtering
• Explicit workflow agent IDs added to WORKFLOW_AGENT_LIST
• Agent prompt updated to enforce correct skill workflow: list_skills → load_skill → follow loaded skill

Also thanks to @bytecii for fixing a related issue where agents were trying to use skills before loading them.

🔗 PR: https://github.com/eigent-ai/eigent/pull/1422 🔗 PR: https://github.com/eigent-ai/eigent/pull/1398

---

🎨 Toolkit Icons Aligned with Tool Categories

Visual clarity for every tool action.

Big thanks to @eureka928 for refactoring toolkit icons to display based on tool type rather than agent type.

What's new:

• 30+ toolkit-to-icon mappings across 7 categories (Dev/Code, Browser/Web, Documents, Media, Communication, Integrations, Meta)
• Icons now visually distinguish Terminal vs Browser vs Note vs File actions at a glance
• Wrench fallback for unknown/MCP toolkits
• Removed duplicate computation in worker node rendering

🔗 PR: https://github.com/eigent-ai/eigent/pull/1218

---

📂 IDE Selection Moved to Inline Dropdown

Less friction, more flow.

Another great contribution from @eureka928 — the "Open in IDE" feature now lives right where you need it.

What's changed:

• Replaced the single icon button + separate Settings dropdown with an inline DropdownMenu in the Folder view
• Three options directly available: Open in VS Code, Open in Cursor, Open in File Manager
• Last choice remembered automatically
• Removed the now-redundant Settings → General IDE preference section

🔗 PR: https://github.com/eigent-ai/eigent/pull/1261

---

🐛 Bug Fixes & Stability

A strong round of bug fixes from across the community.

• macOS Fontations crash fix — Disabled Chromium's unstable Rust-based font engine (fontations_ffi) that caused EXC_BREAKPOINT crashes on Apple Silicon. Thanks @nitpicker55555. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1382

• Browser cookies login fix on Windows — Resolved FileNotFoundError when launching Electron login browser via npx on Windows packaged apps. Thanks @nitpicker55555. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1397

• Model chooser redirect fix — "Choose model" button now correctly navigates to Agents/Models instead of the deprecated Settings/Model page. Thanks @bytecii. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1400

• Duplicate message blocks fix — Resolved duplicate message rendering in chat. Thanks @4pmtong. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1402

• File viewer CSS scoping — CSV/MD renderer styles no longer leak into the rest of the app, fixing typography and layout shifts when switching file types. Thanks @dataCenter430. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1275

• Import path casing fix — Corrected import path casing for WorkspaceMenu and BrowserAgentWorkspace to prevent build issues on case-sensitive file systems. Thanks @4pmtong. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1426

• Worker node UI fixes — Refactored preview layout to grid-based rendering, fixed stray trailing numbers in terminal previews, and added placeholder slots for consistent 2x2 grids. Thanks @Douglasymlai. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1414

• Layout & styling fixes — Fixed home/project panel collapse, top bar spacing, share button visibility, chatbox layout, and dropdown filter font sizes. Thanks @Douglasymlai. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1409 🔗 PR: https://github.com/eigent-ai/eigent/pull/1413 🔗 PR: https://github.com/eigent-ai/eigent/pull/1417 🔗 PR: https://github.com/eigent-ai/eigent/pull/1423

---

📚 Documentation Updates

• Added SambaNova platform documentation. Thanks @fengju0213. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1384

• Updated Skills/Codex documentation with correct video format. Thanks @Douglasymlai. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1387

• Added built-in example skills (code-reviewer, data-analyzer, report-writer). Thanks @4pmtong. 🔗 PR: https://github.com/eigent-ai/eigent/pull/1370

---

❤️ Community Keeps Raising the Bar

This sprint delivered:

• A full Triggers system (schedule + webhook)
• Native LLaMA.cpp local model support
• Search toolkit across all agents
• Memory-efficient streaming media playback
• Smarter skill access control
• Toolkit icons aligned with tool categories
• Inline IDE selection
• A massive wave of bug fixes and polish

Contributors this cycle: @a7m-1st, @it-education-md, @bytecii, @Zephyroam, @fengju0213, @eureka928, @4pmtong, @nitpicker55555, @dataCenter430, @Douglasymlai

Every single one of these improvements came from contributors who cared enough to improve what they use.

That's the power of open source — and the power of open source cowork.

If you've been thinking about contributing triggers, skills, models, UX, agents, tooling, or cowork workflows, this is your moment.

Let's keep building.