Mixus

What is Mixus?

Mixus is an AI agent platform for law firms that drafts redlines, issues lists and deal models, then sends every output to an attorney for approval. Lawyers work with it from email, Microsoft Word or Slack rather than a separate chat window.

The workflow runs in four stages that the company calls send, run, review and improve. A lawyer forwards a matter to an agent address, copies in the reviewers, and receives tracked changes or spreadsheets back in the same thread. Accepted, edited and dismissed suggestions then feed proposed rules into the firm’s playbooks.

Mixus Inc. is based in San Francisco, according to its terms of service. Chief executive Elliot Katz, a former attorney who previously founded the teleoperation company Phantom Auto, told a Stanford CodeX meeting in March 2026 that several AmLaw 20 firms use the platform. The product started in 2024 as a general, no-code agent builder with connectors such as Gmail, Salesforce and Jira, and older directory listings still describe that version. Mixus does not give legal advice, so the reviewing lawyer stays responsible for every result.

Key Features

Playbook-driven redlining is the core of Mixus, because every tracked change cites the firm rule behind it and an attorney approves it before anything leaves.

• AI Redline: Mixus marks up contracts sent as Word, PDF or other files and returns a Word document with tracked changes. Each suggestion carries a comment naming the playbook rule behind it, plus an issues section listing points that need an attorney’s judgement.
• Email-native agents: Lawyers start work by emailing a matter to an agent address and copying in the reviewers who must sign off. Results come back in the same thread, so there is no new interface to learn and little change management for busy teams.
• Deal Modeling: The deal agent turns an emailed term sheet and cap table into an Excel pro forma with live formulas. Mixus says the workbook arrives in under 15 minutes and includes dilution breakdowns, SAFE conversions and price per share checks.
• Firm playbooks: Playbooks store preferred language, fallback positions and flagged terms by client, contract type or practice area. Attorneys can edit a playbook before an agent uses it, and the platform can draft new rules from exemplar documents.
• Review learning: Every accepted, edited or dismissed suggestion is logged as a reviewer signal. Repeated decisions are grouped into proposed rules that a lawyer approves, so the playbook improves without anyone rewriting it by hand.
• Word, Slack and dashboards: A Microsoft Word add-in lets reviewers accept, edit or dismiss suggestions inside the document itself. A Slack integration offers another entry point, and organisation dashboards track usage, approvals and per-run costs in US dollars.

Pros

Attorney control is why firms pick Mixus: agents halt at approval gates, and each change points to a written rule rather than an unexplained model choice.

• Traceable edits: Every tracked change cites the playbook rule that produced it. Reviewing partners can check the reasoning quickly, which makes supervision and quality control easier than with a plain chatbot answer.
• Nothing new to learn: Work starts and ends in email and Word, which lawyers already use all day. The founder has said attorneys pick it up almost immediately, which reduces the training burden on knowledge teams.
• Strong security posture: The trust pages list SOC 2 Type II, ISO 27001:2022, encryption at rest and in transit, and customer-managed keys. Zero data retention with model providers suits firms bound by strict confidentiality duties.
• Choice of models: Mixus runs on an allowlist of models from Anthropic, OpenAI and Google rather than a single vendor. Firms avoid lock-in and can benefit as newer models are validated for redline pipelines.
• Reusable deliverables: Outputs arrive as real Word and Excel files with track changes and live formulas. Associates can send them on to clients or keep editing them, rather than copying text out of a chat window.

Cons

Opaque pricing is the first hurdle with Mixus, since the legal product is sold through demos while the public pricing page still shows an older chat plan.

• No published price: The legal platform offers only demo requests, and its comparison page mentions per-run billing without any figures. Budgeting therefore needs a sales conversation before a pilot can even be scoped.
• Narrower focus: Mixus has moved from a general agent builder to legal work. Teams that wanted it for sales, support or IT automation, as older directory listings describe, should confirm those uses are still supported.
• Confusing legacy pages: The pricing page and parts of the documentation still describe the earlier chat, community and integrations product. New visitors may struggle to tell which features apply to the legal platform today.
• Review stays essential: The terms state that AI outputs may be inaccurate or incomplete and must be checked. Attorneys remain responsible for every redline and model, so time savings depend on disciplined review habits.
• Few named customers: Mixus refers to Am Law 10 and AmLaw 20 firms but names none of them on its site. Independent reviews and case studies are scarce, which makes outside comparison harder for procurement teams.

Who is Using Mixus?

Large law firms are the main Mixus customers, and the company cites Am Law 10 clients and about 50 agents deployed across practices.

• Transactional attorneys: Corporate and commercial lawyers send contracts for first-pass redlines. They review tracked changes and issues lists before sending markups to the other side.
• Venture and M&A teams: Deal lawyers email term sheets and cap tables to the modelling agent. They use the resulting Excel pro forma to check dilution, conversions and wire amounts.
• Knowledge lawyers: Practice support staff build and maintain playbooks for each client or contract type. They approve proposed rules that emerge from repeated reviewer decisions.
• Firm IT and risk teams: Technology and security staff manage keys, access controls and usage dashboards. They rely on the trust documentation during procurement and audits.

Pricing

Mixus sells its legal agent platform by demo and quote, with no published fee for the current product (official site, checked October 2026).

There’s An AI For That lists a paid tier from $20 a month and Futurepedia mentions a free trial, but both appear to describe the 2025 agent builder rather than the legal platform. The terms add that usage-based features, such as AI token consumption, can cost extra.

Disclaimer: Mixus may set fees per firm and per run, so ask for a written quote that reflects your expected matter volume.

What Makes Mixus Unique?

Email works as the control plane in Mixus: an attorney delegates a matter by forwarding it, and approvals happen by reply rather than inside a new app. Many legal AI tools ask lawyers to open a dedicated workspace or chat panel instead.

The way the playbook grows is the second difference. Mixus records what reviewers accept, edit or dismiss and turns repeated choices into proposed rules, so firm preferences build up over time instead of living in individual prompts. The company frames this human-in-the-loop design as a lesson carried over from its founder’s work on remote supervision of autonomous vehicles.

Compatibilities and Integrations

Mixus connects to email, Microsoft Word and Slack, and it returns standard Word and Excel files that fit existing document habits.

• Email: Matters are forwarded to an agent address with reviewers copied in. Responses and files return in the same thread.
• Microsoft Word: An add-in lets reviewers accept, edit or dismiss suggestions in place.
• Slack: A Slack integration is listed as an extra way to reach the agents.
• AI models: An allowlist covers Anthropic, OpenAI and Google models, with zero data retention by the model provider.
• File formats: Contracts can arrive as Word or PDF files. Outputs are .docx files with tracked changes and .xlsx workbooks with live formulas.

Mixus Tutorials

Getting started with Mixus begins with a demo, after which a typical matter moves from email to approved redline in six steps.

1. Book a demo and agree a pilot scope with the Mixus team.
2. Share exemplar contracts so a playbook can be drafted, then edit its rules.
3. Forward a contract to the agent address and copy in the reviewing attorney.
4. Open the returned Word file and read the tracked changes, comments and issues list.
5. Accept, edit or dismiss each suggestion in Word or by replying to the thread.
6. Approve the proposed playbook rules that emerge from repeated decisions.

How We Rated It

Mixus earns 3.8 out of 5 because its attorney-controlled workflow and security credentials are strong, while pricing and public evidence remain thin.

Summary

Mixus suits law firms that want AI agents for redlines and deal models while keeping every output behind an attorney’s approval. Its main limitation is transparency, because the legal product has no published price and few named customers. The pivot from a general agent builder also leaves legacy pages that can confuse new visitors. Responsibility for advice stays with the reviewing lawyer, not the software. If you trial it, run one contract type through a pilot and track how often reviewers accept suggestions unchanged.