Ode with Anthropic vs Railblocks for VC and PE AI
Eric Friedman Co-founder, Railblocks
Short answer: Ode with Anthropic builds custom systems on Claude for mid sized companies across industries, and it’s a strong choice if you’ve decided to standardize on Claude. Railblocks builds AI operating systems for investment firms, works across model providers, and sets everything up so your firm owns it. The question is whether you want depth in one ecosystem or room to change your mind later.
We should say up front that we’re not neutral. We love Anthropic and use its models. We also think a firm should be able to pick something else without losing what it has built.
What is Ode with Anthropic?
Ode is a services business. Anthropic’s May announcement described engineers working with its Applied AI team to find opportunities, build custom systems on Claude, and support mid sized companies over time, with examples that include manufacturers and healthcare groups. Anthropic’s announcement
It launched under the Ode with Anthropic name in July, backed by Anthropic, Blackstone, Hellman & Friedman, and other investors. Its foundation includes Fractional AI, acquired in May. Ode’s launch announcement
Anthropic’s S-1 describes Ode building on the Claude Platform and supporting customers long term. Anthropic S-1, page 194 If your company is committed to Claude, that’s an advantage: you get implementation help closely tied to the people building the model.
How Ode and Railblocks differ
The biggest difference is who we’re built for. Ode serves mid sized companies across industries. Railblocks focuses on investment firms: PE funds, VC firms, and family offices. Running a fund is a different job from running a portfolio company, even though the two overlap. Both of us use forward deployed engineers, who embed with a client and turn what it needs into working systems. Railblocks 2.0
| Ode with Anthropic | Railblocks | |
|---|---|---|
| Who it serves | Mid sized companies across industries | Investment firms: PE, VC and family offices |
| Models | Built on Claude | Across providers, including Claude |
| Biggest strength | Close ties to the team building Claude | Focus on how investment firms work, plus freedom to choose models |
| Tradeoff | Your direction is tied to one ecosystem | Supporting more than one provider takes more testing and upkeep |
| Best fit | A firm standardizing on Claude | A firm that wants to own its system and keep its options open |
This comparison comes from public information and our own view. We haven’t seen Ode’s contracts, and nothing we’ve read says it prohibits other models.
Why model choice comes up
Often the choice has been made before an implementation team ever shows up. Your firm may have cloud commitments it wants to use. A portfolio company may have an approved stack. Procurement may care most about price, and regional or data requirements can narrow things further.
Our job is to work within all of that. If Claude is the right fit, we use Claude. If you need something else, the work still needs to get done.
Whoever you talk to, including us, ask what they prefer to build on, what commercial relationships they have, and why they recommend what they recommend.
What it means to own your firm’s memory
A firm’s memory is more than its documents. It’s how you define a portfolio metric, why you passed on a deal, which source backs up a number, who can see what, and how the thesis changed between the first investment and the latest board meeting.
Models should be able to use all of that without becoming the only place it lives. Take portfolio reporting: your company records, definitions, source history, and review rules are what last, and the model helps read them and draft commentary. When a better model comes along, you shouldn’t have to rebuild your reporting history to use it. You will want to test the new model on your workflows first, the same way you’d check any change that touches your numbers.
At Railblocks, you own it: the data, the code, and the memory, running on infrastructure you control. We build on Cloudflare and route work across model providers through its AI Gateway. Agents run best on Cloudflare
Using Claude doesn’t mean giving up control
To be fair to Anthropic, using its models doesn’t mean handing over your firm. Its S-1 describes an Agent SDK that runs on a customer’s own compute, along with enterprise permissions and audit controls. Anthropic S-1, pages 201 to 203
So it helps to ask four separate questions of any partner:
- Model choice: Which providers can the system use?
- Ownership: Who owns the data, the code, and what the system learns?
- Hosting: Where does the system run, and where does the model processing happen?
- Control: Who manages access, approvals, and changes?
How to decide
Start with one recurring workflow and a clear idea of a good result. For portfolio reporting, that might be numbers backed by sources, consistent definitions, a clear approval step, and less time reconciling. Measure quality, turnaround, review time, and total cost.
Then ask the practical questions. Can you export your records and decision history? Could another engineer pick up the system? If you want flexibility on models, can your workflow pass the same tests on a different one? Who responds when a connector breaks? What do you get if you change partners?
Get the answers in the agreement. If a standard product already solves the problem, buy that instead.
Frequently asked questions
What is Ode with Anthropic?
Ode is an AI services company backed by Anthropic and major investors. It brings Claude into companies’ operations through custom implementation and ongoing support. Ode
When should a VC or PE firm choose a partner tied to one platform?
When you’ve decided to standardize on that platform and you value deep expertise in it more than choice across providers.
Does working across models mean avoiding Anthropic?
No. Many firms will use Claude heavily. It just means Claude is a choice rather than a requirement.
Does owning the data make a system portable?
It’s the first step. Portability also takes usable exports, your workflow code, integrations, and permissions, which is why we make sure you own the whole system.
Can we switch model providers without extra work?
Expect some testing and adjustment. Because your firm’s knowledge lives outside the model, you keep it all when you switch.
Does every firm need an outside AI partner?
No. Some firms have the engineering capacity in house. A partner helps when you need more engineering depth, investment firm experience, or someone to keep things running.
How we can help
We build systems around how investment firms work, using the best tools from Anthropic and others, and you own what we build. You can see what we build and what it costs.
Bring us one workflow, your current stack, and the constraints that matter. Reach out.