Demonstration only. Corollary is a non-commercial research demo — no warranty, no support, not for commercial use or clinical decisions. Read the disclaimer

All comparisons

Compare · AI research agents

Corollary vs Kosmos and the Edison platform

Kosmos is Edison Scientific's AI scientist, offered by application to R&D teams; Edison also runs the agents that began on the FutureHouse platform. Corollary is open to anyone, free for non-commercial research, and built to run the computation itself.

The short answer

Edison Scientific is the company spun out of the nonprofit FutureHouse. Its flagship, Kosmos, is an AI scientist that reads the literature and analyses data over long runs and writes up what it found; the current version is offered by application, without a published price. The Edison Playground — the Literature, Analysis and Precedent agents that began on the FutureHouse platform as Crow, Finch and Owl — is open to everyone, with a small free monthly allowance.

Corollary is aimed at a different part of the work. Edison advises against using Kosmos for raw-data processing such as read mapping; on Corollary that is the everyday case — aligners, assemblers, nf-core pipelines and 145 open scientific models, run for you, with no hardware of your own.

Choose Kosmos for literature-heavy, hypothesis-level research under enterprise terms. Choose Corollary when the job is processing and analysing data and running models, and you want to start today, free.

Side by side

What it is
CorollaryA research agent with a Linux machine of its own, and a library of open scientific models, programs, databases and pipelines it runs for you.
KosmosKosmos is Edison Scientific's AI scientist for R&D. Edison, spun out of the nonprofit FutureHouse, also runs the agents that began on the FutureHouse platform.[1][6]
How you use it
CorollaryIn the browser, with a free account or signed out. Nothing to install. Also a developer API and an MCP server.
KosmosThe current Kosmos is by application, rolled out as capacity allows. The Edison Playground — Literature, Analysis, Precedent — is open to all, on the web and through an API.[3][8]
Price
CorollaryFree. 30,000 credits a month with an account, 1,000 signed out; database queries never use credits.
KosmosKosmos prices are not published. Every account gets 10 free credits a month for the Playground agents.[3]
Commercial use
CorollaryNot permitted. Corollary is a non-commercial research demonstration, with no warranty or support.
KosmosPermitted. Edison sells enterprise deployments, including in the customer's own environment.[1]
Where the work runs
CorollaryOn our servers: a Linux machine per conversation, and jobs of up to 32 cores, 256 GiB of memory and 24 hours for long runs.
KosmosEdison's cloud: sandboxed containers, some with GPUs.[4]
Scientific AI models
Corollary301 models in the library, 145 of them run here — GPU included, with no setup and no key of your own.
KosmosKosmos can provision GPUs for work such as AlphaFold and docking. No catalogue of runnable models is published.[3]
Programs and pipelines
Corollary43 bioinformatics programs installed or run as jobs, and 143 nf-core pipelines.
KosmosAnalysis writes Python, R and Bash in a Jupyter notebook, inside a container of common packages. Edison advises against using Kosmos to process raw data such as mapping sequencing reads.[7][5]
Databases
Corollary53 public databases queried live, with no API key of your own.
KosmosA literature agent, and Springer Nature journal content licensed into Kosmos.[3][9]
Language model
CorollaryChosen and paid for by Corollary. No API key of your own.
KosmosNot disclosed. Edison describes Kosmos as model-agnostic.[1]
Open source
CorollaryNo. The service is closed source; the models, programs and pipelines it runs are open.
KosmosNo. FutureHouse's earlier Robin system is open source.[11]
Your files
CorollaryUploaded to our servers and kept in the conversation's workspace. Some of our processors are in the United States.
KosmosDatasets of up to 5 GB. Edison states SOC 2 Type II; uploaded data is not intentionally used for training, but other data may be unless you opt out.[5][1][10]

Kosmos's column is from its own public pages, read on 24 October 2026; the numbers link to them below. Corollary's figures are counted from the platform itself and each one links to the page that lists them.

Where Kosmos is the better choice

  • Long, autonomous research runs that read the literature and analyse data together — about 1,500 papers a run at launch — ending in a report where every statement cites code or a paper.[2][12]
  • Licensed journal content from Springer Nature inside Kosmos.[9]
  • Enterprise deployments, including in your own environment, with SOC 2 Type II.[1]
  • A free literature agent, and a Python API for the Playground agents.[3][8]

Where Corollary is the better choice

  • Open now, and free: 30,000 credits a month, with no application.
  • Built for raw data: alignment, assembly, variant calling and 143 nf-core pipelines run on machines of their own.
  • 145 open scientific models run here, each with a page saying what it does and its licence.

Try the same work on Corollary

Ask in plain language. The agent works on a Linux machine of its own: it installs and runs what the job needs, keeps every file, and reads the results back with the numbers explained.

Questions

Is Corollary a free alternative to Kosmos?

Corollary is free: 30,000 credits a month with an account, and database queries never use credits. It is a non-commercial research demonstration, so it is not an alternative for commercial work — for that, Kosmos or one of the other options below may be the right choice.

Is FutureHouse the same as Edison Scientific?

Edison Scientific is the company spun out of FutureHouse in November 2025. FutureHouse remains a nonprofit doing basic research; its public platform and agents moved to Edison, and platform.futurehouse.org now redirects there.

How much does Kosmos cost?

Edison no longer publishes a price. The launch price per run went when founding subscriptions ended in May 2026, and the current Kosmos is offered by application.

Can Kosmos process raw sequencing data?

Edison's own guidance advises against using Kosmos for raw-data processing such as mapping sequencing reads. On Corollary that is routine: the agent runs aligners, assemblers and nf-core pipelines on machines of its own.

Sources

  1. [1]Edison Scientific — home
  2. [2]Edison — announcing Kosmos (5 November 2025)
  3. [3]Edison — Incyte partnership and Kosmos for R&D teams (19 May 2026)
  4. [4]Edison — how we built Kosmos (18 May 2026)
  5. [5]Edison docs — best practices for Kosmos
  6. [6]FutureHouse — announcing Edison Scientific (5 November 2025)
  7. [7]Edison — introducing Edison Analysis (20 November 2025)
  8. [8]Edison docs — Python client
  9. [9]Edison — agreement with Springer Nature (24 September 2026)
  10. [10]Edison — privacy policy
  11. [11]Robin (GitHub)
  12. [12]Kosmos: an AI scientist for autonomous discovery (arXiv)

Written by Corollary Labs, which makes Corollary — read it with that in mind. Everything said about Kosmos comes from the sources above as they read on 24 October 2026; products change, and if a line here is out of date, write to hello@corollary-labs.com and we will correct it. Kosmos and the other names here are trademarks of their owners, used only to identify their products. Corollary is a non-commercial research demonstrator. It is provided free of charge, without warranty or support, and must not be used for commercial purposes or relied upon for clinical, diagnostic, regulatory or safety-critical decisions.

Other Kosmos alternatives