Is Deskferry open source?
No. Deskferry is a hosted product, and its pages publish no source repository and no self-host option. Kortix is the open-source alternative.
A Deskferry alternative comes down to one question: where do the rules live? Deskferry keeps them in its workspace and does the setup for you. Kortix is the open-source AI Operating System: your agents, their skills, your company memory, and every connector in one git repo you own, with the agents working on real cloud computers.
Deskferry cells quote its own pages. Prices were read in October 2026.
Recommended
KortixIf a marketing team wants a working agent this week, Deskferry is a fair answer, and its own page states the offer in three sentences: "DeskFerry gives you AI employees: AI agents that each own a recurring job, like replying to new leads or chasing overdue invoices. They work in Gmail, Slack, HubSpot and 1,500+ other apps, and check with you before anything important" (Deskferry).
Three things in that offer are worth paying for. The reach is broad, and it is the kind a small team cannot build: Gmail, Slack, Notion, HubSpot, Salesforce, Zendesk and the long tail behind them. The approval step is written down as a rule, so a buyer can hold the product to it: "Anything that sends, spends or changes a record waits on your Desk for a yes." And the plans do not meter agents: every tier runs unlimited agents. What changes is credits and workspaces. Growth is $49 a month for 5,000 credits, Pro $99 for 12,000, Scale $299 for 40,000, and the 7-day trial does not convert on its own (pricing).
The models come with the product. "Choose models from OpenAI, Anthropic, Google and xAI, with no API keys to manage," the home page says. For a marketing lead with no procurement path to a model provider, that is a real saving in time.
Compare feature lists and the two products look close. Put them beside each other on one question, who can read and change the rules the agent follows, and they stop being versions of the same thing.
A workspace you configure against a repo you own. On Deskferry the agent, its skills and the app connections are objects in the workspace. That suits a team that wants to click, and it is also why a copy of those definitions is not sitting in your repository. On Kortix they are files. Agent behavior, "description, model, mode, temperature, permission, the prompt itself", "lives entirely in that .md's frontmatter + body". The agents: block in kortix.yaml carries governance only: "connectors/secrets/kortix_permissions/skills/workspace/enabled" (the starter manifest). A marketing lead can open the file that tells the agent how to write, diff it against last quarter, and hand the same file to a new hire.
A vendor's model menu against your own keys. Deskferry switches models per task with no key to hold, and the model calls come out of a vendor credit allowance inside the plan. Kortix takes the other route: "Any model provider with your own keys." Model use goes to the provider, and the ChatGPT plan a team already pays for can serve model calls. The trade is honest in both directions. Deskferry removes a decision and a bill. Kortix keeps both, in your name.
A log you read against a change request you approve. Deskferry records every run, and its activity log is a good answer to the question of what happened overnight: "Nothing fails quietly." A Kortix draft arrives differently. "Nothing merges itself: work reaches main through a change request a person approves, unless an admin granted the merge capability." The reviewer sees the lines that changed. The approval leaves a record with a name on it. For a team whose marketing copy is reviewed by legal or by a founder, that difference decides the purchase.
Take the job most teams start with: a Monday digest that reads the week's campaign numbers and drafts the note for the team.
On Deskferry you build it in the workspace, point your apps at it, and set a schedule. On Kortix it is four edits in one repo, and the commands are the quickstart's own:
curl -fsSL https://kortix.com/install | bashkortix init marketing-deskkortix shipkortix sessions new --prompt "Draft the Monday campaign digest"kortix cr lskortix cr diff 1kortix cr merge 1kortix init writes the project from the starter and kortix ship pushes it, connecting the declared apps as it goes. The session works on its own branch, and the last three commands list, read and merge the change request it opens (Read the docs).
The weekly part is a trigger in kortix.yaml, an entry under triggers: that carries its type, schedule, agent and prompt on that one entry:
triggers: - slug: monday-digest type: cron cron: "0 0 8 * * 1" timezone: America/New_York agent: kortix prompt: | Read last week's campaign numbers and draft the Monday digest.The skill file is the second piece, and it is the one a marketer writes. "Your voice, your claims and your banned words live in the repo as a skill file every session reads" (the marketing page). The digest's structure, its tone and the numbers it has to cite sit in skills/ beside the agent. That is what "a skill is the markdown that encodes how your company does one job" means in practice (what a skill is). The connector grant is the third: one line in the agents: block, in the repo, reviewed like any other change. The fourth is the review itself, and it is the one that costs a team something. Someone has to read the diff on Monday morning before the note goes anywhere.
That review is the point, and it is also the cost. A team that will not staff it should not buy this shape of product.
Plenty of teams should. If nobody on the marketing side wants to open a terminal, Deskferry's marketplace and no-key setup will get a first agent running in an afternoon. Its 1,500+ app catalog is also wider than what most teams will connect in a year.
Stay on Deskferry if the workspace is where you want the definitions to live. Stay if an all-in monthly price, models included, suits your finance team better than model keys and a credit pool. Stay if your approval step is one person saying yes on a queue rather than a diff someone has to read.
A repo fits where the claim list itself has to be provable. A regulated claim list, a voice guide that legal signed off, or a review that has to show what changed wants a file with a history and an approval attached to it. What a voice file in the repo looks like is the same argument at the level of one skill. The platform comparison puts Deskferry beside the other products a buyer meets first. The monthly bill is where the two shapes separate again. A credit plan charges for work done. A self-hosted repo charges for the server and the models.
The answers are the vendors' own words, read in October 2026.
No. Deskferry is a hosted product, and its pages publish no source repository and no self-host option. Kortix is the open-source alternative.
$49 a month for 5,000 credits, $99 for 12,000 and $299 for 40,000, with unlimited agents on every plan and a 7-day trial. The credits are what a month of agent work draws down.
Yes, if the team will run its own copy. Kortix is $0 with 500 credits a month on your own provider keys or ChatGPT subscription, then credits from $50 a month, and self-host is free on a server you already pay for.
Deskferry, no: it publishes no self-host option. Kortix, yes. Run it on Kortix Cloud, in your VPC, or on your own on-prem network.
The agent, the skill and the approval are files you can read tonight, and the repo travels with you.