Have us deliver a workflow as a scoped engagement — or license the Akribes stack and have your team build on it. The two paths are priced differently, and the rest of this page is honest about which factors drive which.
A scoped, mostly-once-and-done engagement. We scope it with you, gather examples from your domain experts, generate and iterate the workflow, and hand it off as a typed service your code can call through the SDK.
Single workflow, one body of examples, one deployment environment.
10–20 days
Teams evaluating whether examples-driven workflows fit the work they actually do. Flat-priced by design.
Multi-workflow, production SLA, full audit trails.
2–4 weeks to first workflow live
Teams running AI in customer-facing systems. Volume discounts apply.
Isolated tenant or customer-controlled cloud, 7-year audit retention, SOC-2 Type II evidence.
3–5 weeks with compliance review
Finance, healthcare, legal, and regulated government. Includes on-prem option.
A small workflow with deep verification on a dedicated tenant can cost more than a high-volume workflow with light verification on shared infrastructure.
How many examples you supply, how often they're refreshed, and how diverse the edge cases are. Larger and more varied example sets drive longer iteration cycles.
Number of steps, branches, and tool calls in the generated workflow. A single-pass classification costs less than a multi-step research-and-summarize pipeline.
Shared multi-tenant, dedicated single-tenant, or customer-controlled cloud. Isolation and compliance requirements drive infrastructure cost.
Fast, standard, or full. Fast runs a quick consistency check; standard adds source retrieval; full adds multi-source cross-checking. Most customers mix per workflow.
90 days (standard), 1 year (extended), or 7 years (regulated-industry). Longer retention means more verified-evidence storage and replay capacity.
Services engagements bundle model-provider token costs on a published schedule. License pricing handles inference differently — see Path B.
Podesta holds the API key and absorbs all model-provider token costs.
Provider tokens pass through at 50% below the provider's published rate.
Tokens pass through at the provider's current market rate.
The Akribes server, Studio, and the SDKs licensed to your team. Build as many workflows as you want; integrate them into your product through the SDK; we run the platform underneath. Ongoing, not once-and-done.
Workflow runs your end users trigger through your product. The platform fee scales with that volume — not with the number of workflows you build.
License terms scale with the revenue your product generates from workflows running on Akribes. We agree on the shape up front, in plain numbers, so a successful product doesn't reopen the contract.
Managed cloud, dedicated tenant, or your own cloud / on-prem with signed container images. Isolation and compliance review drive cost.
Shared channel, business-hours response with SLA, or 24/7 on-call.
Build as many as you want. No per-workflow fee, no fixture limits — adding a workflow doesn't change your bill.
Your test set is yours. No per-case fee, no dataset cap. The eval system is supposed to be used.
Benchmarks are just Akribes scripts. Their runs either count as normal executions — same rate as anything else — or come out of an unmetered pool on plans tuned for running evals at scale. We don't meter eval runs separately.
Either bring your own provider tokens (every major model provider supported), or have Podesta proxy them at a transparent +0.5% surcharge so you don't have to set up and top up three-plus provider accounts.
Tell us about the product you're building. We come back with terms.
In a services engagement, Podesta delivers the workflow end-to-end — we scope it with you, gather examples, build, iterate to passing, and hand off a workflow your developers call from code through the SDK. In a license, your team uses the same software in-house: your domain experts upload examples, generate workflows, and integrate via the SDK; we provide the platform, training, and support. Both produce the same artifact — a tested, SDK-callable workflow.
No. Both paths involve scoped planning time, which rules out self-serve free tiers. If you want to evaluate Podesta with a real workflow, start with a Proof-of-Concept engagement under the services path.
Yes. Proof-of-Concept engagements run 10–20 days with one workflow and one body of examples. They're flat-priced and answer one concrete question: does an examples-driven workflow fit the work you actually do?
Different on each path. Services engagement: through Dec 31, 2026 Podesta holds the API keys and pays OpenAI, Anthropic, and Google directly — token costs are bundled into the engagement fee. From January 2027 through Q2 2027, provider tokens pass through at 50% below the provider's published rate. Starting Q3 2027, tokens pass through at the provider's current market rate. The platform fee (workflow generation, iteration, audit) is separate in every phase. License: either bring your own provider tokens (no markup), or have Podesta proxy them for a transparent +0.5% surcharge so you don't have to manage three-plus provider accounts.
Yes, on Standard Production and Dedicated Infrastructure engagements. Proof-of-Concept engagements are flat-priced by design — the point is a clean evaluation, not a pricing negotiation.
Yes, under the Dedicated Infrastructure shape. Podesta ships signed container images, the workflow runtime, and Helm charts; your operations team can run them in your cloud account or on-prem. Audit and compliance evidence stay with the platform. We are happy to walk through the deployment with your infra team and align with the patterns you already run.
Proof-of-Concept: 10–20 days to a production-shaped result on one workflow. Standard Production: 2–4 weeks to first workflow live across multiple teams. Dedicated Infrastructure: 3–5 weeks depending on compliance review.
Yes. USD is the default quoted currency; EUR, GBP, and CHF are available for Standard Production, Dedicated Infrastructure, and license agreements. Proof-of-Concept engagements quote in USD.