Apollo is the one API behind the Colossal Capital ecosystem. Akashic's code analysis, Atlas's document intelligence, and every in-house agent route through it. You can use it the same way they do, over REST or MCP, without adopting anything else we make.
Everything you need to build production AI applications
Chat, completion, and structured analysis with model routing handled for you. Defaults are sensible; per-request overrides exist when you disagree.
A Neo4j knowledge graph and Weaviate hybrid search (BM25 plus vectors) sit behind retrieval endpoints, so agent answers ground in your documents and code rather than model memory.
A persona pool (planner, editor, reviewer, tester, kb_curator, skill_curator) that plans, writes, and reviews work with graph context. Personas are editable per tenant.
Telemetry from your runs becomes SFT and DPO data for LoRA fine-tunes, evaluated against baseline before anything is promoted.
Every inference request is scored on five axes before it's placed: cost, availability, model quality for the task, context window fit, and time-to-available. The winner gets the request; the decision is logged so you can audit why.
| Provider | When the router picks it |
|---|---|
| Together AI | The default for speed. Runs the ATLAS Turbo configurations, roughly a 4x speedup on our workloads. |
| HuggingFace TGI | On Kubernetes inside your VPC when data isn't allowed to leave. Zero egress. |
| HuggingFace API | Fallback when primaries are saturated. |
| OpenRouter | Burst capacity for spiky workloads. |
| RunPod | GPU on demand for jobs that want a whole card. |
// connect any MCP client { "mcpServers": { "apollo": { "command": "npx", "args": ["apollo-mcp-server"], "env": { "APOLLO_API_URL": "https://apollo-api.colossalcapital.co", "APOLLO_API_KEY": "ak_..." } } } } // tools appear in Claude, Cursor, Windsurf, Cline: // run_analysis, kg_query, semantic_search, generate_tickets, repo_agent, ...
Each agent is provisioned through Auth0 with its own machine identity and OAuth-scoped access to your GitHub, Linear, Jira, and Slack. No shared service accounts.
Outbound requests carry a Cloudflare Signed Agents profile (RFC 9421), so third parties can verify which agent made a request, and you can revoke one identity without disturbing the others.
Agent domains currently in use across the ecosystem:
By default, training is per-tenant and private: your repos, tickets, and conventions are tuned into LoRA adapters that only your account can load. No co-mingling.
If you opt in, anonymized policy traces can also feed the shared master model, which applies policy transfer and reward shaping across tenants and is only promoted when it beats baseline. Opting in earns ecosystem credits; opting out later leaves your private adapters untouched.
Some personas have matured into products in their own right: compliance review, trading sentiment, legal extraction, code review — each with its own adapter, eval suite, and MCP tool surface.
Added to any Colossal Capital tier. The tier (free and up) sets API limits, training budget, and team size; Apollo is the add-on that turns on the AI backend. Akashic is included with it.
Get started