Key facts
| Agent stack | Chat completions, function calling and embeddings are live |
| Use cases | Research synthesis, proposal assembly, interview summaries, deliverable QA |
| Isolation | Scoped keys and separate projects or deployments per engagement |
| Retrieval | plugsky-embed plus RAG over approved engagement material |
| Models | 30+ models; small aliases for summarising, stronger ones for synthesis |
| Deployment | Plugsky cloud, your VPC, on-prem and air-gapped options |
| Pricing model | Flat monthly self-serve plans with unlimited fair-use usage |
| Roadmap | Assistants and batch endpoints are coming soon |
TL;DR
- Automate research synthesis and assembly, not judgement or client advice.
- Isolate every engagement with its own key and index.
- Require citations so a consultant can verify any claim before reuse.
- Keep partner review for anything that reaches the client.
- Measure proposal cycle time and rework, not output volume.
How it works, step by step
- Choose one deliverable type such as a research memo or proposal first draft.
- Create a separate key, project and retrieval index per engagement.
- Index only approved material and require citations on every factual claim.
- Use tools for template retrieval, prior-deliverable search and data lookups.
- Route drafts to a partner or manager review before client delivery.
- Log sources and model versions, then measure cycle time and revision count.
Try it yourself
Open the AI agent prompt generator →
Where consulting agents earn their keep
Consulting is a research-and-document business, which fits bounded agent loops:
- Research synthesis: combine market data, filings and interview notes into a structured findings memo.
- Proposal assembly: pull approved methodology, bios and case studies into a first draft.
- Interview summaries: turn transcripts into themes with attributed quotes.
- Deliverable QA: check a draft against the approved checklist before partner review.
Judgement, recommendations and client advice stay with the consultant.
Architecture and engagement isolation
Isolation is the design constraint, not model choice:
- Per-engagement keys: separate credentials and projects so one client's data cannot leak into another's run.
- Per-engagement index: retrieve only from documents approved for that client.
- Tools: template retrieval, prior-deliverable search and internal data lookups, all read-only.
- Composer: drafts with citations and a clear list of assumptions.
- Reviewer: a partner or manager signs off before anything leaves the firm.
Where a client contract demands more, deploy a dedicated VPC or on-prem instance for that engagement.
Confidentiality, retention and records
Client contracts often specify where data can be processed and how long it may be retained. Configure the region and retention before the first document is indexed, scope keys to the engagement, and use audit logs to produce a record of who asked what and when. Keep prompts and evaluation sets versioned so quality changes are explainable, and store the model version with each output. Because the API is compatible, the same agent code runs in the cloud, a VPC or on-prem without rewriting, which matters when clients sit under different constraints.
Pilot and measurement
Prototype on plugsky-micro or plugsky-lite on one internal, non-client corpus, then use the 14-day full-access trial on a real deliverable with the client's permission. Track first-draft cycle time, partner revision count, research hours saved and citation quality. Self-serve plans are flat monthly with fair-use usage, so engagement volume does not create billing swings. Roll out one deliverable type at a time, keep prompts in version control, and re-test whenever templates or model versions change.
Honest comparison
| Capability | Plugsky | Typical agent framework | Building in-house |
|---|---|---|---|
| Tool calling | Live function calling over approved internal sources | Framework-specific adapters | You build and maintain |
| Engagement isolation | Scoped keys and separate projects or deployments | Shared tenancy by default | You design tenancy |
| Deployment | Cloud, VPC, on-prem and air-gapped | Usually cloud-only | You operate GPUs and serving |
| Pricing | Flat monthly self-serve, fair-use usage | Varies by framework and model | GPU plus operations cost |
| Model choice | 30+ models behind one API | Varies by provider | You host every model |
Frequently asked questions
Can the agent deliver work to clients directly?
No. It prepares drafts with citations; a partner or manager reviews and owns anything that reaches the client.
Which endpoints do consulting agents use?
Chat completions, function calling, JSON mode and embeddings are live. Assistants and batch endpoints are coming soon.
How do we keep engagements separate?
Use scoped keys, separate projects and separate retrieval indexes per engagement, with a dedicated VPC or on-prem instance where contracts require it.
Is there a free plan?
Yes. The free plan includes plugsky-micro and plugsky-lite with no credit card, suitable for internal prototypes and evaluation sets.
How does pricing work?
Self-serve plans are flat monthly with unlimited fair-use usage. See the live pricing page for current plans and enterprise options.
How do we handle client confidentiality terms?
Set the processing region and retention before indexing anything, keep keys scoped, and log access for auditability.
How should we measure success?
Track first-draft cycle time, partner revision count, research hours saved and citation quality on real deliverables.