Eigent
Pricing
A cream facade with blue windows refracted through fine vertical glass

AI workspace

  • OverviewSee the whole workspace.
  • Spaces and SessionsKeep each project's files and work together.
  • Skills & ConnectorsTeach agents your way of working. Connect your tools.
  • AutomationsRun Tasks on a schedule or a trigger.
  • Remote ControlCheck in from your browser or phone.
  • Use casesSee real work, finished.

Updates

  • ChangelogCatch up on the latest releases.
  • RoadmapFollow what comes next.
  • WindowsDownload Eigent for Windows.
  • MacDownload Eigent for macOS.
  • LinuxDownload Eigent for Linux.
  • Open SourceInspect, adapt, and make it yours.
  • EnterpriseRoll Eigent out across your company.
Concrete balconies and golden railings refracted through fine vertical glass

Enterprise

  • OverviewAn AI workspace for every team, in your environment.
  • MonitoringSee every run, step by step.
  • GovernanceSet who can build, run, and approve.
  • ControlLimit what agents can reach.
  • Security and trustHow we protect your data.
  • Enterprise servicesWe build it with you, then hand it over to your team.
  • Talk to salesPlan a proof of concept for your company.
A sunlit facade and geometric shadows refracted through fine vertical glass

Environments

  • OverviewTrain and test agents on real work.
  • Environment catalogBrowse 30 categories of environments.
  • Request an environmentTell us what your agents need to learn.
  • AgenticRL-BenchCan agents do agentic RL? See the leaderboard. (opens in a new tab)
  • Env-LabCustom RL environments, training data, and benchmarks for real work. (opens in a new tab)
Weathered window frames refracted through fine vertical glass

Learn

  • DocsGet started and go deeper.
  • BlogIdeas, releases, and stories.
  • SupportFind answers and get help.

Company

  • AboutMeet the people behind Eigent.
  • CareersJoin the team building Eigent.

Community

  • CAMEL-AIThe multi-agent research community behind Eigent. (opens in a new tab)

AI workspace

  • OverviewSee the whole workspace.
  • Spaces and SessionsKeep each project's files and work together.
  • Skills & ConnectorsTeach agents your way of working. Connect your tools.
  • AutomationsRun Tasks on a schedule or a trigger.
  • Remote ControlCheck in from your browser or phone.
  • Use casesSee real work, finished.

Updates

  • ChangelogCatch up on the latest releases.
  • RoadmapFollow what comes next.
  • WindowsDownload Eigent for Windows.
  • MacDownload Eigent for macOS.
  • LinuxDownload Eigent for Linux.
  • Open SourceInspect, adapt, and make it yours.
  • EnterpriseRoll Eigent out across your company.

Enterprise

  • OverviewAn AI workspace for every team, in your environment.
  • MonitoringSee every run, step by step.
  • GovernanceSet who can build, run, and approve.
  • ControlLimit what agents can reach.
  • Security and trustHow we protect your data.
  • Enterprise servicesWe build it with you, then hand it over to your team.
  • Talk to salesPlan a proof of concept for your company.

Environments

  • OverviewTrain and test agents on real work.
  • Environment catalogBrowse 30 categories of environments.
  • Request an environmentTell us what your agents need to learn.
  • AgenticRL-BenchCan agents do agentic RL? See the leaderboard. (opens in a new tab)
  • Env-LabCustom RL environments, training data, and benchmarks for real work. (opens in a new tab)

Learn

  • DocsGet started and go deeper.
  • BlogIdeas, releases, and stories.
  • SupportFind answers and get help.

Company

  • AboutMeet the people behind Eigent.
  • CareersJoin the team building Eigent.

Community

  • CAMEL-AIThe multi-agent research community behind Eigent. (opens in a new tab)
Pricing
Blogs
Produto|Sep 25, 2026

Eigent Release Notes v1.0.5: Session Recovery, Task Queues & Better Previews

Pick what runs next, return to interrupted work, and inspect the processes and files behind a Session

Douglas LaiDouglas Lai
Share to
Eigent Release Notes v1.0.5: Session Recovery, Task Queues & Better Previews
  • 🔄 Return to an Interrupted Session
  • 📋 Choose What Runs Next
  • 🖥️ See the Processes Behind the Work
  • 🧩 Configure a Space with the Resources You Already Have
  • ✨ More Model Choice, Clearer Model Errors
  • ⚙️ Smaller Fixes That Keep Work Moving
  • 🔁 Upgrade Notes
  • ❤️ More Control Over Work in Progress
Automate Everything with
AI Workforce on Desktop
Download Eigent

Eigent v1.0.5 makes ongoing work easier to control and easier to return to. Interrupted Sessions recover their conversation and Task state more consistently. A redesigned queue lets you decide which Task should run next, while the Session side panel gives you a clearer view of background processes and their output.

This release also brings configured Skills, Connectors, MCP servers, and models closer to each Space, and adds Meta Model API and Muse Spark. The common thread is clarity: what is running, what finished, what is waiting, and which configuration a Session will use.


🔄 Return to an Interrupted Session

Losing a connection or restarting the app should not make completed work look unfinished.

v1.0.5 improves how Eigent reconciles a Run after a stream interruption and reconstructs Session history when you come back. Conversation content, Task state, elapsed time, completed output, and file summaries are restored more consistently. Historical durations remain visible for completed Tasks, while timers stop for failed historical Tasks instead of continuing to count.

What's improved:

  • More dependable recovery — interrupted Sessions show the state Eigent has already recorded
  • Completed results stay visible — finished output and file summaries remain attached to the Task that produced them
  • Accurate historical time — completed durations survive recovery and failed Tasks no longer keep a running timer
  • Model continuity — a resumed Session uses its accepted model configuration, including after a usage-limit-driven model switch

This makes it easier to understand what happened before an interruption and continue from the right point.

Thanks to @4pmtong for Session history, Run reconciliation, and model continuity, and @Douglasymlai for interrupted-Session recovery and completed file summaries.

🔗 PRs: #1920, #1927, #1937, #1948


📋 Choose What Runs Next

When several follow-up Tasks are waiting, their order should be yours to control.

The redesigned queue supports drag-and-drop and keyboard reordering. You can choose a waiting Task to run next, with an explicit stop step when another Task is already running. Clearer waiting, stopping, and starting states show where the transition stands instead of making the queue appear to jump ahead.

What's new:

  • Reorder queued Tasks — move work into the order that fits your priorities
  • Keyboard controls — adjust the queue without relying on a pointer
  • Choose the next Task — select the waiting item that should start when the current work stops
  • Clear transitions — see when a Task is waiting, stopping, or starting

Tasks within a Session still run in order. The queue gives you control over that order and makes an intentional handoff visible.

Thanks to @Douglasymlai for the queue redesign.

🔗 PR: https://github.com/eigent-ai/eigent/pull/1934


🖥️ See the Processes Behind the Work

A generated site or a long-running command is easier to review when you can see whether its process is still alive.

The Session side panel now lists background processes associated with the work. From there, you can inspect terminal output, open a local server preview, or stop a process. The file and browser preview paths are sturdier too: PDFs render more reliably, failed or unsupported previews have a recovery path, local browser loads are easier to find, and HTML files up to 10 MiB can be previewed.

What's improved:

  • Background process controls — find running processes, view their output, and stop them from the Session
  • Local server previews — open a running server's page beside the conversation
  • File preview recovery — recover from failed or unsupported previews instead of being left with a dead panel
  • Larger HTML previews — inspect local HTML files up to 10 MiB
  • More reliable PDFs and browser loads — fewer dead ends when reviewing generated work

The result is a more complete review loop: see the Task, inspect what it started, and open the output in the same Session.

Thanks to @Douglasymlai for background process controls and sturdier file and browser previews, and @4pmtong for larger HTML previews.

🔗 PRs: #1940, #1938, #1939, #1942


🧩 Configure a Space with the Resources You Already Have

A Space can now show the Skills, Connectors, MCP servers, and models available to its setup, making it easier to assemble the right working environment.

For a published, set-up Space, its configured default model can be applied when you create a new Session. A model you choose manually for that Session still takes precedence, and existing Sessions retain their own model selections. Recovery also preserves the model accepted when a Session began, so the current global setting does not silently replace it.

What's improved:

  • Resource discovery — browse configured Skills, Connectors, MCP servers, and models from Space settings
  • Space model defaults — new Sessions can start with the model configured for that Space
  • Session choice stays yours — manual selections and existing Session models are preserved
  • Consistent recovery — a resumed Session continues with its accepted model configuration

This keeps the Space's intended setup close to the work while respecting choices already made inside a Session.

Thanks to @4pmtong for Space resource discovery and model defaults.


✨ More Model Choice, Clearer Model Errors

v1.0.5 adds Meta Model API and Muse Spark to custom model settings. It also improves model discovery and the feedback you receive when a provider is unavailable, a key is invalid, access is limited, or credit runs out.

The release includes fixes for reasoning-effort settings, OpenAI and Azure tool requests, Anthropic-compatible tool calls, and OpenAI Workforce initialization with saved credentials. Those changes help a supported model configuration make it from Settings into an actual Task without an avoidable request or tool-schema failure.

Thanks to @neptune-me for Meta Model API and Muse Spark, @Douglasymlai for model setup and error feedback, and @4pmtong for the Anthropic-compatible tool-call fix.

🔗 PRs: #1904, #1936, #1966


⚙️ Smaller Fixes That Keep Work Moving

Several focused fixes make day-to-day work less fragile:

  • Clearer service messages — credit, access, connection, and provider errors are easier to understand, with fewer duplicate reminders
  • Long-running triggers — a server-side time threshold alone no longer marks an active trigger execution as failed
  • Safer workspace operations — better write coordination, provider retries, cancellation, process shutdown, and incremental Memory processing
  • Dependable Space deletion — deletion completes even when the local backend is temporarily unavailable
  • Settings polish — clearer Skills file browsing, less Connector loading flicker, persistent Settings headers, and distinct network proxy Save, Reset, and Restart actions
  • Message feedback — thumbs-up and thumbs-down feedback is connected to message-level events
  • Git path handling — filenames containing spaces, Unicode, or line breaks are handled correctly
  • Windows startup — Git is bundled with Windows x64 builds, so a missing system Git installation does not prevent backend startup
  • Credential file handling — environment files are replaced atomically and use owner-only permissions on supported systems

These are the details that matter when a Task lasts longer than a quick demo or when a workspace contains real files and configuration.

Thanks to @Douglasymlai for service messages, Settings and feedback improvements; @4pmtong for Space deletion, Git and Windows startup fixes, and atomic credential writes; and @BetterAndBetterII for owner-only credential-file permissions.


🔁 Upgrade Notes

Some older interrupted Tasks may need to be started again if their saved model capabilities no longer match the current configuration. The recovery improvements make the state of those Tasks clearer, but they do not assume an incompatible model can resume safely.


❤️ More Control Over Work in Progress

Eigent v1.0.5 delivers:

  • More consistent Session recovery and historical Task results
  • A reorderable queue with clearer stop and start transitions
  • Background process controls and stronger file, PDF, HTML, and browser previews
  • Space resource discovery and model defaults for new Sessions
  • Meta Model API and Muse Spark, alongside model and provider fixes
  • Reliability improvements across triggers, workspaces, Settings, Git paths, Windows startup, and credential files

This release makes a busy workspace easier to follow. You can see what has finished, choose what happens next, inspect the processes behind the output, and return to a Session with a clearer account of its state.

🔗 Release: https://github.com/eigent-ai/eigent/releases/tag/v1.0.5

🔗 Full changelog: https://github.com/eigent-ai/eigent/compare/v1.0.4...v1.0.5

Let's keep building.

Recent Posts

Notas da versão do Eigent v1.0.8 — Nova tela inicial, busca global e conversas mais claras
ProdutoNov 17, 2026

Notas da versão do Eigent v1.0.8 — Nova tela inicial, busca global e conversas mais claras

O Eigent v1.0.8 reformula a tela inicial e o campo de mensagens, adiciona busca entre espaços e exibe por padrão a linha do tempo das ações do agente.

Douglas LaiDouglas Lai
GPT-6.2 Astra e Sol: O Que Há de Novo, Preços e Benchmarks
SetorNov 17, 2026

GPT-6.2 Astra e Sol: O Que Há de Novo, Preços e Benchmarks

GPT-6.2 Astra e Sol são os modelos mais recentes da OpenAI para o ChatGPT Work, Codex e a API. Veja o que mudou, os preços, os benchmarks e qual usar.

EigentEigent
O Que É NVIDIA Nemotron for Science? Modelos Abertos para Agentes de Pesquisa
Nov 17, 2026

O Que É NVIDIA Nemotron for Science? Modelos Abertos para Agentes de Pesquisa

NVIDIA Nemotron for Science explicado: o modelo aberto, o dataset de química LitmusBench v0.2 e receitas de pós-treinamento para construir agentes de IA para pesquisa científica.

EigentEigent
Automate everything with AI workforce on desktop
Download Eigent

Experimente o Eigent hoje

Baixe o app desktop open source e comece a automatizar com uma força de trabalho de IA na sua máquina.

Baixar Eigent
Eigent

Receba as últimas novidades e tutoriais sobre automação de workforce com IA.

Obrigado por se inscrever!

ProdutoEigentAmbientesPreçosEmpresarial
ExplorarSoluçõesCasos de usoHabilidadesPluginsBlogNovidades
DesenvolvedoresDocumentaçãoGitHubCAMEL-AIFundo Open SourceParceiro
BaixarPara código aberto
EmpresaSobre nósMarcaCarreirasSuporteTermos de usoPolítica de privacidadeSegurança e confiançaPolítica de cookiesPolítica de reembolso e teste

Todos os direitos reservados © 2026 EIGENT UK LTD