Kortix vs Multica: Action Approvals or Issue Handoffs?
Compare Kortix and Multica on action permissions, issue ownership, workflow triggers, result checks and hosting choices.
Founder, Task Machine
Kortix vs Multica asks where your team needs the control point: on an agent's connector action or on an issue moving between owners. Kortix's guide documents rules that hold an exact action for approval. Multica's workflow docs describe handing work to a person or agent and sending rejected work back.
Those controls solve different problems for a small software team. A review step can judge a finished change. Permission to use a connected service decides whether an agent may take a particular action while doing the work, before that finished change reaches review.
Kortix keeps shared configuration in files
Kortix's developer documentation describes agents, skills and connector settings configured in repository files. Saved JavaScript workflows coordinate agents through pipelines and parallel calls, with arguments for later runs. This fits a team that wants to express its agent setup and coordination in code.
Its user guide describes My Kortix proposing a multi-agent workflow plan before execution. You approve the plan, and its card shows phases, agent count and progress. Agent work itself lives in sessions.
Experimental teams add a lead, teammates and a shared list of numbered tasks with owners, dependencies and status. They are off by default and last only for the lead's session. Its docs do not establish persistent work records that both people and agents can own with a separate discussion and history, so the team task list should not be treated as an issue tracker.
Multica puts the issue at the center
Multica's issue documentation describes assignments to a person, agent or squad. Discussion, deliverables, status and multiple agent runs stay attached to that issue. For a bug that moves from investigation to implementation and review, the issue remains the shared reference.
Its project workflows define handoffs to agents or people, a next step and a send-back step when review rejects work. A squad leader can route issues to members, and project leads can be people or agents. Those roles suit a team that already thinks in assigned issues and project stages.
Multica can also support work outside engineering, including content and community work. The deciding property is the issue-and-handoff model, not whether the artifact is code. Its workflow docs do not describe general conditional branches or verifier checks for each acceptance requirement.
An approval queue and a review step differ
Kortix's guide documents Allow, Ask and Block rules for connector actions. Review collects held actions, and a person approves or denies the exact call. If the call changes, it needs another approval. For an agent preparing customer outreach, that gives you a published control on the proposed connected action.
Multica's Inbox documentation describes notifications, while workflows can assign a review stage to a person. Its docs do not describe configuring which agent actions need approval, or collecting those action requests in the Inbox. A human review stage remains useful, but it should not be assumed to govern every service action taken before that stage.
Choose a test that exposes the distinction. Ask where an agent must stop before sending a message, then separately ask who reviews the completed report. Kortix publishes an exact-action approval mechanism, while Multica publishes a human handoff for reviewing work. The appropriate choice depends on which stop your team needs to enforce.
Triggers do not start every kind of workflow
Kortix supports schedules and signed webhooks that start agents or wake existing sessions. Its connector events are experimental and off by default. The saved JavaScript workflow tool has another boundary: it starts only from a person's message, not from a trigger, reminder or another session.
Multica's Autopilot guide describes schedules or webhooks that can create an issue or run without one. Project workflows can then move issue work between owners. Existing issues also support recurring wakeups, although regular checks require an end date.
For unattended work, inspect the failure path as carefully as the trigger. A Kortix trigger-started session waits at most 30 minutes for a person and then stops. Multica's Run only Autopilots skip when the runtime is unavailable, and failed runs in that mode do not retry automatically.
A scheduled research agent and a scheduled multi-agent script are separate requirements. If you need a weekly chain that starts without a person, Kortix's published workflow-tool restriction is relevant even though its agents support schedules. If you choose Multica's Run only mode, arrange the runtime and recovery process around its documented skip behavior.
Compare checks, execution and billing honestly
Kortix's experimental Goals use a separate verifier session that did none of the work. It checks every done criterion as pass, fail or inconclusive, and anything short of all passes returns the goal to work. Goals are experimental and off by default, so this should not be presented as an unconditional default for all sessions.
Multica lets issues contain acceptance criteria and lets a review step return rejected work. Its public docs do not establish that delivered results are checked against defined criteria. A checklist in an issue and an independent criterion-by-criterion verifier therefore represent different published mechanisms.
Both offer hosted execution and ways to use your computer, with self-hosting options. Kortix's desktop app can provide filesystem, shell or screen access under controls on that computer. Multica connects a laptop or server with installed coding tools or uses managed cloud sandboxes. Its provider guide makes model availability depend on the installed tool, configuration, signed-in account and subscription entitlements.
Kortix's pricing page lists Free with 500 monthly credits on the first workspace. Credits starts at $50 a month for 5,000 credits shared by unlimited members. Managed models add 5.5%, own-key usage adds 5% above $1,000 a month, and compute and models bill separately.
Multica's homepage says internal organizational use, including self-hosting, is free. External hosting or embedding requires a commercial license, and the homepage does not publish a cloud price. These terms need separate evaluation from the model access and infrastructure your team will use.
Kortix and Multica, point by point
The table gives both products the same questions, with each answer's source and check date. Experimental features and narrower published capabilities remain explicit.
Both columns answer the same questions from each company's public pages. Yes means the company publishes the capability the question asks about, Partly means something narrower, No means it says it does not offer it, and Not published means its public pages do not describe it. Every answer links to its source and shows when we last checked it. The oldest check on this page was November 13, 2026.
| What you compare | Kortix | Multica |
|---|---|---|
| Goals |
Kortix
Yes
Kortix describes agents running a search program that tracks queries, rewrites declining pages and measures results. Its Goals feature assigns an agent to done criteria and scheduled wakes, but is experimental and off by default. |
Multica
Partly
Projects have a shared goal and a member or agent lead. The documentation does not describe linked outcome metrics monitored by that lead. |
| Leads |
Kortix
Yes
An agent team has a lead that splits work, coordinates teammates through a shared task list and combines their answers. Teams are experimental, off by default and exist only for the lead's session rather than as saved teams. |
Multica
Yes
Project leads can be people or agents, and squad leaders route issues to members. The documentation does not describe goal leads that monitor outcomes or project leads that independently review goal conflicts. |
| Tasks |
Kortix
Partly
Agent teams have numbered tasks with owners, status and dependencies, while sessions keep the conversation. Its docs do not establish work records that both people and agents can own with their own discussion or history. |
Multica
Yes
Assign an issue to a person, agent or squad. Its discussion, deliverables, status and multiple agent runs remain on the issue. |
| Recurring work |
Kortix
Yes
Triggers start agents on schedules or signed webhooks, and loops wake existing sessions with their history. Connector events are also available experimentally and are off by default. |
Multica
Yes
Autopilots start from schedules or webhooks, creating an issue each time or running without one. Recurring wakeups also return an agent to an existing issue, but regular checks require an end date. |
| Workflows |
Kortix
Yes
JavaScript workflows coordinate agents through parallel calls and pipelines, and saved scripts can run again with arguments. The workflow tool starts only from a person's message, not a trigger or reminder. |
Multica
Yes
Project workflows hand steps to agents or people, with a next step and a send-back step for rejected work. They do not document general conditional branches or per-requirement verifier checks. |
| Autonomy |
Kortix
Yes
Allow, Ask and Block rules control connector actions, and Review collects held approvals. A person approves or denies the exact action, and a changed call needs a new approval. |
Multica
Partly
An Inbox collects notifications and workflows can hand a review step to a person. The documentation does not describe configuring which agent actions require approval or collecting approval requests in the Inbox. |
| Learned autonomy |
Kortix
Not published
Its public docs describe editable action rules and a reflection agent that improves skills and instructions, but no proposals to change an agent's permissions based on approval history. |
Multica
Not published
Multica documents agent access and workflow handoffs. Its documentation does not describe proposing changes to an agent's autonomy based on approval history. |
| Memory |
Kortix
Yes
Agents keep memory as editable Markdown files in Drive and read it at the start of later tasks. My Kortix saves preferences and corrections, and you can ask what it remembers or tell it to forget something. |
Multica
Partly
Multica retains Hermes memory per agent and can optionally enable Codex native memory. Its documentation does not describe whether users can inspect or correct that memory. |
| Planning |
Kortix
Yes
My Kortix proposes a workflow plan in conversation before running multiple agents. You approve it first, and its card shows phases, agent count and progress. |
Multica
Yes
Chat agents can create issues when asked, and issues can contain acceptance criteria. The documentation does not describe a required plan and acceptance-criteria review before an agent starts. |
| Intake |
Kortix
Yes
Signed webhook triggers route outside events to a chosen agent or existing session. Software Factory can receive issues and failures into a ticket store and give each ticket a worker. |
Multica
Yes
A webhook can trigger an autopilot that creates an issue, and a squad leader can route an assigned issue to members. The documentation does not describe a project-token intake that places a proposal in the Inbox first. |
| Connectors |
Kortix
Yes
Kortix publishes a catalog of 3,000+ apps plus MCP, OpenAPI and raw HTTP connections. You can add custom connectors, and signed webhooks start agent work. |
Multica
Yes
Agents can use individual or shared MCP servers, and autopilots accept webhook triggers. The documentation does not describe the same catalog of API, command-line and MCP connectors. |
| Models |
Kortix
Partly
You choose a workspace default, pin a model to an agent or select one per session. Ordered fallback models handle failures, but its docs do not describe automatic model selection based on each task or step. |
Multica
Partly
You can select a model and thinking level for supported agent runtimes. The documentation does not describe automatic model selection for each work stage based on price and budget. |
| Cloud or your own computer |
Kortix
Yes
Kortix offers its hosted cloud and a self-hosted deployment. You can also connect your own computer through the desktop app for filesystem, shell or screen access, with access controlled on that computer. |
Multica
Yes
Connect a laptop or server to run installed coding agents, or use Multica's managed cloud sandboxes. You can also self-host the Multica service. |
| Your own subscription |
Kortix
Yes
Every plan accepts your own provider keys or a ChatGPT subscription. The guide also documents Claude Pro and Max subscriptions, whose model calls use no Kortix credits. |
Multica
Yes
Multica invokes installed AI coding tools such as Claude Code, Codex and GitHub Copilot CLI using your local sign-in or model-provider setup. Available models depend on the tool, local configuration, signed-in account and subscription entitlements. |
| Pricing |
Kortix
Yes
Free includes 500 monthly credits on the first workspace. Credits starts at $50 a month for 5,000 credits shared by unlimited members. Managed models add 5.5%. Own-key usage adds 5% above $1,000 a month. Compute and models bill separately. |
Multica
Partly
Multica says internal organizational use, including self-hosting, is free, and external hosting or embedding requires a commercial license. It does not publish a cloud price on its home page. |
| Correctness |
Kortix
Yes
Experimental Goals use a separate verifier session that did none of the work. It checks every done criterion as pass, fail or inconclusive, and anything short of all passes sends the goal back to work. |
Multica
Partly
Issues can contain acceptance criteria, and a workflow can send rejected work back from a review step. The documentation does not establish that results are checked against defined criteria. |
For software work
| What you compare | Kortix | Multica |
|---|---|---|
| Code review rules |
Kortix
Partly
Software Factory verifies changes and adds independent review before a merge gate. Its docs do not specify code review against repository rules. |
Multica
Partly
You can assign an agent a review skill or instructions and set a workflow review step. The documentation does not establish that code is reviewed against rules. |
| Adversarial review |
Kortix
Yes
Software Factory puts an independent review after the worker's verification and feedback, before a merge gate. Its docs do not describe a separate high-risk change classification for that review. |
Multica
Partly
A workflow can send work to a review agent and return rejected work for changes. Its documentation does not establish an independent second review of risky changes. |
| Risk rating |
Kortix
Not published
Its developer docs allow risk labels on submitted Review items, but do not describe a rating of a code change before merge. |
Multica
Partly
Linked pull requests show checks and merge conflicts. The documentation does not describe rating the exact revision's merge risk. |
| Pull requests |
Kortix
Yes
Software Factory assigns tickets to workers that verify changes and open pull requests with demos, followed by feedback, independent review and a merge gate. A person merges unless automatic merging is configured. |
Multica
Partly
Agents can open pull requests, and linked issues show their state, checks and conflicts. The documentation does not establish that agents work on pull requests with review feedback. |
A fact we cannot find or cannot check is left out, and prices can change after the date shown. Task Machine publishes this comparison and is an interested party. Kortix and Multica are named only to identify them, and there is no relationship between the companies. Has something changed? Tell us.
Who each one suits
Pick Kortix if repository-based configuration, saved agent scripts and exact connector-action approvals fit your team. Its workflow tool's person-message requirement is a limit for unattended scripted coordination, and its Goals and teams remain experimental.
Pick Multica if people and coding agents need shared issues, project leads and review handoffs. Its docs do not describe per-action approval settings, and Run only Autopilots need a recovery plan for skipped or failed runs.
Where Task Machine fits
Task Machine combines Tasks owned by people or agents with per-agent Ask first or Automatic action settings. Its Workflow builder defines repeated steps, branches, human questions and approvals, with verifier feedback for failed steps. Task reviews check each acceptance requirement before human review, which makes both the action boundary and the result check explicit.
Task Machine is the wrong pick if you need Multica's self-hosting option, because its public setup docs do not describe self-hosting the workspace service. See Switch from Multica to Task Machine. Use Task Machine vs Kortix and Task Machine vs Multica to map one unattended job's trigger, action approval and rejected-result handoff before choosing.