The place where intelligent transformation happens.
Spaceplace brings models, company context, role workspaces and agentic workflows together, so people and AI agents complete work in the systems you already use.
Spaceplace brings models, company context, role workspaces and agentic workflows together, so people and AI agents complete work in the systems you already use.
ask_space · Conversationscrm_list_leadsask_space · Knowledgecrm_move_lead_stageCHANGEcrm_move_lead_stagecrm_list_leads ask_space| # | Question | Score |
|---|---|---|
| 1 | What is our refund policy for a damaged order? | Waiting to run…Agent answering…Judge scoring…Passed |
| 2 | Which support requests are past their deadline? | Waiting to run…Agent answering…Judge scoring…Passed |
| 3 | Summarize this week's calls for the sales manager | Waiting to run…Agent answering…Judge scoring…Needs review |
| 4 | Which clients have not paid their last invoice? | Waiting to run…Agent answering…Judge scoring…Passed |

Spaceplace is built to give modern enterprises advanced AI capabilities in one place, instead of dozens of different tools.
Use your own API keys, models on your own servers, or models hosted in the data centre you choose. For each task, the router uses the model you choose, or the one with the best balance of price and quality.
A hard task, so the strongest model
Long, routine text, so a strong model at a lower cost
A simple question, so the lowest cost
Use your own accounts with Anthropic, OpenAI and other providers.
Models hosted in the region your rules require, such as the EU.
Open models, such as Qwen, run on servers you control.

Run your agents and workflows on your own cases with different models at each step. Evaluation Lab compares quality and cost, and recommends the optimal combination that meets your quality bar.
Each employee works in Space with AI built on their role's context, skills and agents, to speed up everyday work. It is processed securely and kept isolated within your company's workspace.

Here is a draft:
Hello, and sorry your order is late. It leaves our warehouse today and arrives on Thursday. Delivery is on us this time.
Three are ready:
Three things:

Start building right away in our cloud, with your keys and your models. When your security rules require it, deploy Spaceplace inside your own perimeter, on your own inference, so your data and models stay on your infrastructure.
Knowledge, records, your industry's rules, your processes and the history of every conversation and decision, in one place. Every answer shows its sources, and each role reaches only the parts of the context it is allowed to.

Give each employee an AI allowance by role. Stop each agent at a monthly budget and cap what one action may cost. Set a cap for every provider and key, and see spend as it happens.
Turn a process into connected steps. Agents do the routine work in your systems, and a person approves where it matters.

Damaged item · order #4821
crm.find_order #4821Refund · damaged item
intent.skillApproved by you
Order updated in your system
erp.update_order #4821Agents read and act in your CRM, ERP, telephony and messengers: Salesforce, HubSpot, SAP, 1C, Binotel, Telegram, Viber, WhatsApp, Gmail and more. Some connect out of the box, others through their API during setup.

Spaceplace interviews your knowledge keepers, the people who know how the work really gets done, and maps how the process actually runs. From that map it builds an agentic process for you to approve, and each run checks the outcome and informs the next.

«First I ask for a photo, then I check the order.»
Interviewed«We confirm the damage within two days, if someone tells us.»
Interviewed«Anything above $100 needs my sign-off.»
Interviewed«Refunds go out of the ERP once a week.»
InterviewedBegin with one workflow or one role. Each next one reuses the company context, connections, rules and tests already in place, so it starts from what the first one built.

Short answers. The details are on the sub-pages, or on a call.
Talk to an engineer →Models from Anthropic, OpenAI and other providers through your own API keys, models hosted in the region you choose, or open models on your own servers. The router picks one for each task, or you pin the one you want.
Spaceplace runs in our cloud, with your own keys and models. For deployment inside your own perimeter, on your own models and inference, talk to our engineers about your case.
Your team can build agents, workflows and role workspaces, connect your systems and release them on its own, with support and recommendations from our engineers.
Evaluation Lab runs it on your own cases and scores every answer before release, and compares versions, so a new one never does worse than the last.
Each role sees what its permissions allow, and every answer shows its sources. How we protect your data is on the Trust page.
Caps for each provider account and API key, and the cost of every call on record. Ask us about budgets per agent, per task and per employee.

Bring one workflow or one role. We'll show you how Spaceplace runs it, on your systems.