an open-source alternative to the dots, bots and muses of the world, run an office of claude code/codex like agents on your laptop, sandboxes or anywhere, uses your existing subscriptions
HTML agent-harnessagent-orchestrationagentsai-agentsautonomous-agentsclaude-codecodexcoding-agentsdesktop-appdeveloper-toolselectrongemini-cliharnessharness-engineeringlocal-firstmemorymulti-agentopencodeorchestrationtypescript
blog(mission): link Munder Difflin mentions to the download page
chaitanyagiri82309c3fe776
diff --git a/blog/src/posts/claude-code-max-plan-tips.md b/blog/src/posts/claude-code-max-plan-tips.mdindex 1153490a..1aa8513b 100644--- a/blog/src/posts/claude-code-max-plan-tips.md+++ b/blog/src/posts/claude-code-max-plan-tips.md@@ -25,7 +25,7 @@ To get the most out of Claude Max in Claude Code, spend the plan on work, not onPrices are in our post on [what Claude Code costs](/blog/how-much-does-claude-code-cost/), so this one sticks to using the plan well. If you already run more than one terminal, read [how to manage multiple Claude Code sessions](/blog/manage-multiple-claude-code-sessions/) too, because every extra session draws from the same allowance.-You can do all of this by hand, or use [Munder Difflin](https://munderdiffl.in/), a free and open source desktop app that runs your Claude Code agents on the login you already have. Give each agent its own model and token cap, schedule long jobs for when you are away, and watch every session from one screen. One tip first: in 0.5.2 the default agent model is Fable 5, which on Max can use at most half your weekly limit, so pick another default under Settings > Agents & Models.+You can do all of this by hand, or use [Munder Difflin](https://harnessmd.com/download), a free and open source desktop app that runs your Claude Code agents on the login you already have. Give each agent its own model and token cap, schedule long jobs for when you are away, and watch every session from one screen. One tip first: in 0.5.2 the default agent model is Fable 5, which on Max can use at most half your weekly limit, so pick another default under Settings > Agents & Models.## How do Claude Max usage limits work in Claude Code?diff --git a/blog/src/posts/codex-max-plan-tips.md b/blog/src/posts/codex-max-plan-tips.mdindex d5514da7..615cfe49 100644--- a/blog/src/posts/codex-max-plan-tips.md+++ b/blog/src/posts/codex-max-plan-tips.md@@ -25,7 +25,7 @@ OpenAI has no plan called Codex Max. The plan people mean is ChatGPT Pro, whichThree things muddy the name: Anthropic's Claude Max plan, OpenAI's older GPT-5.1-Codex-Max model, and Codex's **Max** reasoning setting, which gives a model more time on one task ([OpenAI, Models](https://learn.chatgpt.com/docs/models)). None is a Codex plan. Still choosing between tools? Our [Codex CLI vs Claude Code comparison](/blog/codex-cli-vs-claude-code/) covers that. The habits in [managing several Claude Code sessions](/blog/manage-multiple-claude-code-sessions/), like one scope per session, carry straight over to Codex.-You can apply every tip below by hand, or use [Munder Difflin](https://munderdiffl.in/), a free and open source desktop app that runs Codex agents next to Claude Code and other CLIs. Each Codex agent signs in with your own Codex login, and you pick its model per agent, from Luna for routine edits to Astra for the hard bug. Our [mixed engine office guide](/blog/run-a-mixed-engine-office/) shows a layout.+You can apply every tip below by hand, or use [Munder Difflin](https://harnessmd.com/download), a free and open source desktop app that runs Codex agents next to Claude Code and other CLIs. Each Codex agent signs in with your own Codex login, and you pick its model per agent, from Luna for routine edits to Astra for the hard bug. Our [mixed engine office guide](/blog/run-a-mixed-engine-office/) shows a layout.## Is there a Codex max plan?diff --git a/blog/src/posts/grok-bot-alternatives.md b/blog/src/posts/grok-bot-alternatives.mdindex 6d2ca9b8..ba4d4dac 100644--- a/blog/src/posts/grok-bot-alternatives.md+++ b/blog/src/posts/grok-bot-alternatives.md@@ -57,7 +57,7 @@ Six good options, grouped by why you're looking.### You want coding agents on your own machine-Munder Difflin, which we make, is a free and open source desktop app for a narrower need: a team of coding agents (Claude Code, Codex, Gemini CLI and nine others) working on your own computer. A Slack message can start an agent, missions run on a schedule, and with your own OpenAI API key you direct the agents by voice, with a spoken confirmation before anything destructive. It will not clear your inbox, so for errands and email, pick from the other groups.+[Munder Difflin](https://harnessmd.com/download), which we make, is a free and open source desktop app for a narrower need: a team of coding agents (Claude Code, Codex, Gemini CLI and nine others) working on your own computer. A Slack message can start an agent, missions run on a schedule, and with your own OpenAI API key you direct the agents by voice, with a spoken confirmation before anything destructive. It will not clear your inbox, so for errands and email, pick from the other groups.### Your work lives in files on your computerdiff --git a/blog/src/posts/herdr-alternatives.md b/blog/src/posts/herdr-alternatives.mdindex 5335616b..54e5e058 100644--- a/blog/src/posts/herdr-alternatives.md+++ b/blog/src/posts/herdr-alternatives.md@@ -65,7 +65,7 @@ What herdr adds is knowing which agent needs you right now. Herdr's [concepts doSeven are active, each checked on its own repo or site on 14 Sep 2026. Pick by your reason for leaving.* **Claude Squad, for something smaller on top of the tmux you already have.** [Claude Squad](https://github.com/smtg-ai/claude-squad) is a terminal UI for Claude Code, Codex, OpenCode and Amp that gives each agent its own tmux session and git worktree. It needs tmux and the GitHub CLI, is AGPL 3.0 licensed, and shipped v1.0.20 on 20 Aug 2026.-* **Munder Difflin, for agents that work as a team.** This one is ours: a free and open source desktop app for macOS, Windows and Linux where each agent is a real CLI in a real terminal, from any of twelve CLIs. Michael, your clone, hands out the work, the agents share a long-term memory, and anything waiting on you lands on the ASK ME board. The trade: quitting the app stops every agent terminal, so there is no detach and walk away.+* **[Munder Difflin](https://harnessmd.com/download), for agents that work as a team.** This one is ours: a free and open source desktop app for macOS, Windows and Linux where each agent is a real CLI in a real terminal, from any of twelve CLIs. Michael, your clone, hands out the work, the agents share a long-term memory, and anything waiting on you lands on the ASK ME board. The trade: quitting the app stops every agent terminal, so there is no detach and walk away.* **cmux, for a native Mac window.** cmux is a Ghostty based macOS terminal with vertical tabs and notification rings for coding agents. It is Mac only, so no help on a headless Linux box.* **Orca, for a full GUI on any desktop.** Orca is an MIT licensed app for macOS, Windows and Linux that runs each agent in its own worktree, with a built-in browser and SSH remote worktrees. It shipped v1.4.201 on 13 Sep 2026, and our [Orca vs Munder Difflin](/blog/orca-vs-munder-difflin/) post goes further.* **Conductor, if agents in the cloud suit you.** Conductor is a Mac app that runs Claude Code, Codex, Cursor and OpenCode agents, and its homepage now leads with Conductor Cloud, where each agent gets an isolated microVM. It was on 0.85.0 as of 9 Sep 2026.diff --git a/blog/src/posts/omniagent-alternatives.md b/blog/src/posts/omniagent-alternatives.mdindex a298aa59..e8dc2d2d 100644--- a/blog/src/posts/omniagent-alternatives.md+++ b/blog/src/posts/omniagent-alternatives.md@@ -49,7 +49,7 @@ The setup is more than one download. The README lists Python 3.12 or newer, `uv`Everything goes through a server. Even solo, you run one on your laptop and every client attaches to it. That is what makes shared sessions possible, and it is more moving parts than one person at one desk needs. Our piece on [local first vs cloud agent SDKs](/blog/local-first-vs-cloud-agent-sdks/) covers that trade in general.-If a server is more than you need, Munder Difflin is the local version of the idea: a free and open source desktop app for macOS, Windows and Linux where every agent is a real CLI in its own terminal, across twelve providers including Claude Code, Codex, Gemini CLI, Copilot and Cursor. Michael, your clone, takes your brief and hands out the work, agents share memory, and per agent token caps plus a circuit breaker keep spend in check. It has no server wide policy layer like Omnigent's, and the agents run only while your computer is on.+If a server is more than you need, [Munder Difflin](https://harnessmd.com/download) is the local version of the idea: a free and open source desktop app for macOS, Windows and Linux where every agent is a real CLI in its own terminal, across twelve providers including Claude Code, Codex, Gemini CLI, Copilot and Cursor. Michael, your clone, takes your brief and hands out the work, agents share memory, and per agent token caps plus a circuit breaker keep spend in check. It has no server wide policy layer like Omnigent's, and the agents run only while your computer is on.Governance is written as code. Omnigent policies return allow, deny or ask, and they stack at three levels: server wide for admins, per agent in the YAML, and per session in the UI, with session rules checked first. Built-in handlers ask before file and shell tool calls, cap tool calls per session and block expensive models past a spend limit, with optional warnings before that. You can toggle the built-ins in the UI, but a rule beyond them means writing a Python handler, which is either a relief or a chore depending on who owns Python on your team. On the managed service, [Databricks' docs](https://docs.databricks.com/aws/en/omnigent/) (updated 11 Sep 2026) say only the built-in contextual policies are supported.@@ -62,7 +62,7 @@ Five are worth a look, each checked on its own repo or docs on 14 Sep 2026.| Alternative | What it is | Fits you if || --- | --- | --- || Claude Code agent teams | Built into Claude Code, experimental, switched on with `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1`. A lead session spawns teammates that share a task list and message each other ([docs](https://code.claude.com/docs/en/agent-teams)). | You only run Claude and want nothing new installed. It will not mix in Codex or Cursor. |-| Munder Difflin | Our free and open source desktop app, described above. | You want many CLIs working as one team on your own machine. |+| [Munder Difflin](https://harnessmd.com/download) | Our free and open source desktop app, described above. | You want many CLIs working as one team on your own machine. || [Orca](https://github.com/stablyai/orca) | MIT licensed desktop app for macOS, Windows and Linux that runs Codex, Claude Code, OpenCode or Pi side by side, each in its own git worktree, with a mobile companion. | You want parallel agents in a GUI and you compare their results yourself. || Paperclip | MIT licensed, self hosted Node server and React UI that runs agents as a company, with an org chart, monthly budgets per agent and approval for new hires ([repo](https://github.com/paperclipai/paperclip)). | Your governance question is who reports to whom and what each agent may spend, rather than rules on every tool call. || LangGraph, OpenAI Agents SDK, Claude Agent SDK | Libraries you build agents on. | You are shipping an agent product, not supervising coding agents. See [frameworks vs a local harness](/blog/crewai-autogen-vs-a-local-agent-harness/). |diff --git a/docs/blog/claude-code-max-plan-tips/index.html b/docs/blog/claude-code-max-plan-tips/index.htmlindex ab93fb42..260dc51c 100644--- a/docs/blog/claude-code-max-plan-tips/index.html+++ b/docs/blog/claude-code-max-plan-tips/index.html@@ -170,7 +170,7 @@<div class="prose"><p>To get the most out of Claude Max in Claude Code, spend the plan on work, not on rereading context. Pick the model per task, <code>/clear</code> between tasks, plan before big edits, keep sessions inside the one hour cache, and watch <code>/usage</code> so the weekly limit never surprises you. Anthropic’s figures below were checked on 14 Sep 2026.</p><p>Prices are in our post on <a href="/blog/how-much-does-claude-code-cost/">what Claude Code costs</a>, so this one sticks to using the plan well. If you already run more than one terminal, read <a href="/blog/manage-multiple-claude-code-sessions/">how to manage multiple Claude Code sessions</a> too, because every extra session draws from the same allowance.</p>-<p>You can do all of this by hand, or use <a href="https://munderdiffl.in/">Munder Difflin</a>, a free and open source desktop app that runs your Claude Code agents on the login you already have. Give each agent its own model and token cap, schedule long jobs for when you are away, and watch every session from one screen. One tip first: in 0.5.2 the default agent model is Fable 5, which on Max can use at most half your weekly limit, so pick another default under Settings > Agents & Models.</p>+<p>You can do all of this by hand, or use <a href="https://harnessmd.com/download">Munder Difflin</a>, a free and open source desktop app that runs your Claude Code agents on the login you already have. Give each agent its own model and token cap, schedule long jobs for when you are away, and watch every session from one screen. One tip first: in 0.5.2 the default agent model is Fable 5, which on Max can use at most half your weekly limit, so pick another default under Settings > Agents & Models.</p><h2 id="how-do-claude-max-usage-limits-work-in-claude-code" tabindex="-1">How do Claude Max usage limits work in Claude Code? <a class="anchor" href="#how-do-claude-max-usage-limits-work-in-claude-code" aria-hidden="true">#</a></h2><p>Two meters run at once: a session limit that resets every five hours, and a weekly limit. The weekly one covers all models and resets at a fixed time assigned to your account (<a href="https://support.claude.com/en/articles/11049741-what-is-the-max-plan">Anthropic, What is the Max plan</a>, checked 14 Sep 2026). Max 5x and Max 20x multiply Pro’s usage per session, and chat, Claude Code and the IDE extensions share one pool. Because both meters are shared across models, switching with <code>/model</code> does not bring back a spent session or week. The separate Opus and Sonnet family limits are different: switch outside that family and you keep working (Claude Code errors reference). Run <code>/usage</code> to see both bars and their reset times.</p><figure class="fig fig-inline"><img src="/blog/assets/media/claude-code-max-plan-tips/note-1.png" alt="An hourglass beside a seven day strip with some days filled in" loading="lazy" decoding="async" /></figure>diff --git a/docs/blog/codex-max-plan-tips/index.html b/docs/blog/codex-max-plan-tips/index.htmlindex 93150b18..1c69bd3a 100644--- a/docs/blog/codex-max-plan-tips/index.html+++ b/docs/blog/codex-max-plan-tips/index.html@@ -170,7 +170,7 @@<div class="prose"><p>OpenAI has no plan called Codex Max. The plan people mean is ChatGPT Pro, which comes as Pro 5x at $100 a month or Pro 20x at $200 (OpenAI pricing, checked 14 Sep 2026). To make the most of it, run the smallest model and effort that works, keep one chat per task, and check <code>/status</code> before big jobs.</p><p>Three things muddy the name: Anthropic’s Claude Max plan, OpenAI’s older GPT-5.1-Codex-Max model, and Codex’s <strong>Max</strong> reasoning setting, which gives a model more time on one task (<a href="https://learn.chatgpt.com/docs/models">OpenAI, Models</a>). None is a Codex plan. Still choosing between tools? Our <a href="/blog/codex-cli-vs-claude-code/">Codex CLI vs Claude Code comparison</a> covers that. The habits in <a href="/blog/manage-multiple-claude-code-sessions/">managing several Claude Code sessions</a>, like one scope per session, carry straight over to Codex.</p>-<p>You can apply every tip below by hand, or use <a href="https://munderdiffl.in/">Munder Difflin</a>, a free and open source desktop app that runs Codex agents next to Claude Code and other CLIs. Each Codex agent signs in with your own Codex login, and you pick its model per agent, from Luna for routine edits to Astra for the hard bug. Our <a href="/blog/run-a-mixed-engine-office/">mixed engine office guide</a> shows a layout.</p>+<p>You can apply every tip below by hand, or use <a href="https://harnessmd.com/download">Munder Difflin</a>, a free and open source desktop app that runs Codex agents next to Claude Code and other CLIs. Each Codex agent signs in with your own Codex login, and you pick its model per agent, from Luna for routine edits to Astra for the hard bug. Our <a href="/blog/run-a-mixed-engine-office/">mixed engine office guide</a> shows a layout.</p><h2 id="is-there-a-codex-max-plan" tabindex="-1">Is there a Codex max plan? <a class="anchor" href="#is-there-a-codex-max-plan" aria-hidden="true">#</a></h2><p>No, but ChatGPT Pro is the closest thing, and its bigger tier just got harder to buy. OpenAI split Pro in two on 9 Apr 2026, adding a $100 tier with 5x the Codex usage of Plus and a launch boost to up to 10x that ran through 31 May 2026 (<a href="https://x.com/OpenAI/status/2042295688323875316">OpenAI on X</a>), so that boost is over.</p><p>Pro $200, labelled Pro 20x, is still the highest usage tier, but since 10 Sep 2026 OpenAI has paused new sign-ups and upgrades to it (<a href="https://help.openai.com/en/articles/9793128">About ChatGPT Pro tiers</a>). Existing subscriptions renew as usual. If you have Pro 20x, think before you cancel or downgrade: once it lapses, you cannot buy it back until the pause lifts.</p>diff --git a/docs/blog/grok-bot-alternatives/index.html b/docs/blog/grok-bot-alternatives/index.htmlindex 24db02d0..add405c4 100644--- a/docs/blog/grok-bot-alternatives/index.html+++ b/docs/blog/grok-bot-alternatives/index.html@@ -187,7 +187,7 @@<p><strong>ChatGPT Work</strong> is OpenAI’s agent inside ChatGPT, announced on 9 Jul 2026. Plugins connect it to Slack, Microsoft Teams, Google Drive, email and calendars, Scheduled Tasks run on a timer or when an event happens, and you decide when it needs your approval. In the Mac and Windows desktop app it can also use your local files, and there OpenAI offers Work on every plan, Free included.</p><p><strong>Gemini Spark</strong> suits you if your life runs on Gmail, Drive and Calendar. Google’s page says it keeps working with your phone and laptop off and is designed to check with you before major actions. It is offered to Google AI Pro and Ultra subscribers in select countries, plus select business users.</p><h3 id="you-want-coding-agents-on-your-own-machine" tabindex="-1">You want coding agents on your own machine <a class="anchor" href="#you-want-coding-agents-on-your-own-machine" aria-hidden="true">#</a></h3>-<p>Munder Difflin, which we make, is a free and open source desktop app for a narrower need: a team of coding agents (Claude Code, Codex, Gemini CLI and nine others) working on your own computer. A Slack message can start an agent, missions run on a schedule, and with your own OpenAI API key you direct the agents by voice, with a spoken confirmation before anything destructive. It will not clear your inbox, so for errands and email, pick from the other groups.</p>+<p><a href="https://harnessmd.com/download">Munder Difflin</a>, which we make, is a free and open source desktop app for a narrower need: a team of coding agents (Claude Code, Codex, Gemini CLI and nine others) working on your own computer. A Slack message can start an agent, missions run on a schedule, and with your own OpenAI API key you direct the agents by voice, with a spoken confirmation before anything destructive. It will not clear your inbox, so for errands and email, pick from the other groups.</p><h3 id="your-work-lives-in-files-on-your-computer" tabindex="-1">Your work lives in files on your computer <a class="anchor" href="#your-work-lives-in-files-on-your-computer" aria-hidden="true">#</a></h3><p><strong><a href="https://support.claude.com/en/articles/13345190-get-started-with-cowork">Claude Cowork</a></strong> needs a paid Claude plan. In the macOS or Windows desktop app, Claude works in the folders you choose, and Anthropic says it can’t reach anything else and asks before deleting. By default the work, local files included, runs on Anthropic’s servers (in beta), so it is closer to Grok Bot than it looks, with a tighter fence around your files.</p><h3 id="you-want-the-memory-logins-and-model-on-your-side" tabindex="-1">You want the memory, logins and model on your side <a class="anchor" href="#you-want-the-memory-logins-and-model-on-your-side" aria-hidden="true">#</a></h3>diff --git a/docs/blog/herdr-alternatives/index.html b/docs/blog/herdr-alternatives/index.htmlindex d130ed5a..50796dbd 100644--- a/docs/blog/herdr-alternatives/index.html+++ b/docs/blog/herdr-alternatives/index.html@@ -196,7 +196,7 @@ herdr-windows-x86_64.zip<p>Seven are active, each checked on its own repo or site on 14 Sep 2026. Pick by your reason for leaving.</p><ul><li><strong>Claude Squad, for something smaller on top of the tmux you already have.</strong> <a href="https://github.com/smtg-ai/claude-squad">Claude Squad</a> is a terminal UI for Claude Code, Codex, OpenCode and Amp that gives each agent its own tmux session and git worktree. It needs tmux and the GitHub CLI, is AGPL 3.0 licensed, and shipped v1.0.20 on 20 Aug 2026.</li>-<li><strong>Munder Difflin, for agents that work as a team.</strong> This one is ours: a free and open source desktop app for macOS, Windows and Linux where each agent is a real CLI in a real terminal, from any of twelve CLIs. Michael, your clone, hands out the work, the agents share a long-term memory, and anything waiting on you lands on the ASK ME board. The trade: quitting the app stops every agent terminal, so there is no detach and walk away.</li>+<li><strong><a href="https://harnessmd.com/download">Munder Difflin</a>, for agents that work as a team.</strong> This one is ours: a free and open source desktop app for macOS, Windows and Linux where each agent is a real CLI in a real terminal, from any of twelve CLIs. Michael, your clone, hands out the work, the agents share a long-term memory, and anything waiting on you lands on the ASK ME board. The trade: quitting the app stops every agent terminal, so there is no detach and walk away.</li><li><strong>cmux, for a native Mac window.</strong> cmux is a Ghostty based macOS terminal with vertical tabs and notification rings for coding agents. It is Mac only, so no help on a headless Linux box.</li><li><strong>Orca, for a full GUI on any desktop.</strong> Orca is an MIT licensed app for macOS, Windows and Linux that runs each agent in its own worktree, with a built-in browser and SSH remote worktrees. It shipped v1.4.201 on 13 Sep 2026, and our <a href="/blog/orca-vs-munder-difflin/">Orca vs Munder Difflin</a> post goes further.</li><li><strong>Conductor, if agents in the cloud suit you.</strong> Conductor is a Mac app that runs Claude Code, Codex, Cursor and OpenCode agents, and its homepage now leads with Conductor Cloud, where each agent gets an isolated microVM. It was on 0.85.0 as of 9 Sep 2026.</li>diff --git a/docs/blog/omniagent-alternatives/index.html b/docs/blog/omniagent-alternatives/index.htmlindex 1f88cc72..13b0b475 100644--- a/docs/blog/omniagent-alternatives/index.html+++ b/docs/blog/omniagent-alternatives/index.html@@ -186,7 +186,7 @@ v0.13.0<p>Four things can push you elsewhere: the setup, the server at the centre, the style of governance, or a wish to write the orchestration themselves. None of them makes Omnigent a bad tool.</p><p>The setup is more than one download. The README lists Python 3.12 or newer, <code>uv</code>, <code>git</code>, Node 22 with <code>npm</code> and <code>pnpm</code>, <code>tmux</code> for the native Claude and Codex wrappers, and <code>bubblewrap</code> on Linux, though the one line installer offers uv, tmux and bubblewrap. On Windows it runs “in a degraded mode”, without the native terminal wrappers or the filesystem and network sandbox. The project still wears an alpha badge, and releases went from v0.9.0 on 11 Aug to v0.13.0 on 9 Sep 2026, so expect things to move under you.</p><p>Everything goes through a server. Even solo, you run one on your laptop and every client attaches to it. That is what makes shared sessions possible, and it is more moving parts than one person at one desk needs. Our piece on <a href="/blog/local-first-vs-cloud-agent-sdks/">local first vs cloud agent SDKs</a> covers that trade in general.</p>-<p>If a server is more than you need, Munder Difflin is the local version of the idea: a free and open source desktop app for macOS, Windows and Linux where every agent is a real CLI in its own terminal, across twelve providers including Claude Code, Codex, Gemini CLI, Copilot and Cursor. Michael, your clone, takes your brief and hands out the work, agents share memory, and per agent token caps plus a circuit breaker keep spend in check. It has no server wide policy layer like Omnigent’s, and the agents run only while your computer is on.</p>+<p>If a server is more than you need, <a href="https://harnessmd.com/download">Munder Difflin</a> is the local version of the idea: a free and open source desktop app for macOS, Windows and Linux where every agent is a real CLI in its own terminal, across twelve providers including Claude Code, Codex, Gemini CLI, Copilot and Cursor. Michael, your clone, takes your brief and hands out the work, agents share memory, and per agent token caps plus a circuit breaker keep spend in check. It has no server wide policy layer like Omnigent’s, and the agents run only while your computer is on.</p><p>Governance is written as code. Omnigent policies return allow, deny or ask, and they stack at three levels: server wide for admins, per agent in the YAML, and per session in the UI, with session rules checked first. Built-in handlers ask before file and shell tool calls, cap tool calls per session and block expensive models past a spend limit, with optional warnings before that. You can toggle the built-ins in the UI, but a rule beyond them means writing a Python handler, which is either a relief or a chore depending on who owns Python on your team. On the managed service, <a href="https://docs.databricks.com/aws/en/omnigent/">Databricks’ docs</a> (updated 11 Sep 2026) say only the built-in contextual policies are supported.</p><figure class="fig fig-inline"><img src="/blog/assets/media/omniagent-alternatives/note-2.png" alt="Hand drawn sketch: a todo, doing, done board with task cards" loading="lazy" decoding="async" /></figure><h2 id="which-omnigent-alternatives-are-worth-trying" tabindex="-1">Which Omnigent alternatives are worth trying? <a class="anchor" href="#which-omnigent-alternatives-are-worth-trying" aria-hidden="true">#</a></h2>@@ -207,7 +207,7 @@ v0.13.0<td>You only run Claude and want nothing new installed. It will not mix in Codex or Cursor.</td></tr><tr>-<td>Munder Difflin</td>+<td><a href="https://harnessmd.com/download">Munder Difflin</a></td><td>Our free and open source desktop app, described above.</td><td>You want many CLIs working as one team on your own machine.</td></tr>