Blackbox AI
Also known as: Blackbox, CyberCoder, Chairman LLM, Blackbox Router, k-agents
Frontier inference platform with coding agents: an OpenAI compatible inference API and router over 300+ models, plus an agentic CLI and cloud agents that run Blackbox, Claude Code, Codex or Gemini on a repository, deployable from multi tenant SaaS to fully air gapped.
Blackbox AI, founded in 2019 and based in San Francisco, sells frontier inference and the coding agents built on it to organizations that need high trust AI. Its Agents API, included on the enterprise plan, sends coding agents to a repository and returns their work as a pull request. A multi agent task runs two to five agents, including Blackbox's own agent, Claude Code, Codex and Gemini, on the same task in parallel so their approaches and results can be compared, and Chairman LLM evaluates the implementations.
An agentic command line tool reads a repository's conventions and handles coding, debugging, setup and deployment from the terminal. Its plan mode lets a person approve or rewrite the plan before work starts and can require approval for each file, and it connects the customer's own MCP servers and runs hooks around each tool call. Skills are written and shared through version control, and a VS Code extension brings the agent into the editor.
Underneath, the Blackbox Router exposes more than 300 open and closed models through one OpenAI compatible endpoint with provider routing, prompt caching and cost aware routing, and Enterprise Inference runs an open weight model the customer chooses on reserved capacity.
The platform is built for regulated customers. Prompts are encrypted before they leave the developer's machine, zero data retention is enforced at the gateway, and PII can be stripped before prompts reach a closed model. Deployment runs from multi tenant SaaS through a dedicated VPC in AWS, Azure or GCP to on premises and air gapped installs, pinned to US or EU regions. Enterprise adds SAML SSO, SCIM, fine grained RBAC over repositories, models and agent capabilities, and audit logs streamed to a SIEM. The site no longer mentions a SOC 2 or ISO 27001 audit. Pricing is enterprise only, on committed annual spend.
Vendor details
Canonical URL
https://blackbox.ai
Category
Coding agent
Subcategory
Coding — assistant
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
The Agents API reaches any GitHub repository over HTTP to create runs, stream logs and open pull requests, and a remote MCP server at agent.blackbox.ai lets the CLI coordinate cloud agents. The Blackbox Router exposes 300+ open and closed models through one OpenAI compatible endpoint with provider routing, and audit logs stream to Splunk, Datadog, Sumo Logic and Elastic. Dedicated VPC deployments run in AWS, Azure or GCP.
In practice
You want to try several coding agents on the same task before trusting one. A Blackbox multi agent task runs its own agent, Claude Code, Codex and Gemini in parallel on your repository so you can compare their approaches, commits and results.
Your platform team wants to start coding agents from its own tools. The Agents API creates runs against any repository over HTTP, streams live logs and opens pull requests with clean commits.
Your code can't leave your network or region. Blackbox deploys to a dedicated VPC, on premises or fully air gapped, pinned to US or EU regions, with zero data retention and prompts encrypted before they leave the machine.
Sources & related URLs
Related / legacy domains
Agentic Index coverage score
8.5 / 14 capabilities · 61%
| Integrations & Tool Calling | Full |
|---|---|
|
The CLI connects the customer's own MCP servers through a config file and runs hooks before and after each tool call, so the agent acts through tools the customer adds. Cloud agents work against GitHub repositories named per task and open pull requests, and audit logs stream to Splunk, Datadog, Sumo Logic and Elastic. Sourceblackbox.ai/cliread 2026-10-25 |
|
| Workflow Orchestration | Full |
|
Blackbox runs several agents (its own, Claude Code, Codex or Gemini, each with a chosen model) on the same repository task in parallel and has Chairman LLM judge the implementations, and the CLI orchestrates sub tasks and parallel tasks across cloud agents. Sourceblackbox.ai/agents and blackbox.ai/cliread 2026-10-25 |
|
| Knowledge Grounding & RAG | Partial |
|
The CLI reads the conventions of the repository, the VS Code agent analyzes the whole project context, and skills stored in the workspace add team knowledge. Blackbox describes no indexing, embedding or retrieval over the customer's code or documents. Sourceblackbox.ai, docs.blackbox.ai CLI and VS Code pagesread 2026-09-29 |
|
| Human Oversight & Guardrails | Full |
|
The CLI's plan mode lets a person approve, comment on or rewrite the plan before work starts and can require manual approval for each file before changes apply. Admins also constrain agents through RBAC over repositories, models and agent capabilities, model allow listing and per workspace content policies, and cloud agent work lands as pull requests. Sourceblackbox.ai/cliread 2026-10-25 |
|
| Security, Identity & Governance | Partial |
|
The enterprise plan includes SAML SSO, SCIM, role based access control and audit logs, zero data retention applies, and PII is stripped before prompts reach closed models. The site no longer mentions SOC 2 or ISO 27001, and its trust page shows no attestation. Sourceblackbox.ai/pricing and blackbox.airead 2026-10-25 |
|
| Observability & Auditability | Partial |
|
The Agents API streams live run logs, and a multi agent task records each agent's status, commits, result and errors. Audit logs stream to a customer SIEM. Blackbox does not say what the logs and audit events contain or whether they record the agent's tool calls and reasoning step by step. Sourceblackbox.ai/agents, blackbox.ai/enterprise and docs.blackbox.ai/api-reference/multi-agent-taskread 2026-09-29 |
|
| Memory & State Persistence | Not documented |
|
A run can be continued as a conversation, which keeps state inside one task. Blackbox describes no agent memory that carries across tasks. Retention windows set how long data is kept, and instruction files hold instructions. Sourceblackbox.ai/enterprise and docs.blackbox.airead 2026-09-29 |
|
| Deployment & Data Residency | Full |
|
Deployment comes in four modes, from multi tenant SaaS and a dedicated VPC in AWS, Azure or GCP to on premises in the customer's datacenter and air gapped with no outbound internet. Data can be pinned to US or EU regions in every mode. Sourceblackbox.ai/enterpriseread 2026-09-29 |
|
| Prebuilt Agents, Templates & Packs | Not documented |
|
CLI skills are written by the customer, and Blackbox publishes no prebuilt agent library, template set or bundled skills. The agents a task can run (Blackbox's own, Claude Code, Codex, Gemini) are interchangeable coding agents doing the same job. Sourceblackbox.ai/agents and docs.blackbox.ai/features/blackbox-cli/skillsread 2026-09-29 |
|
| Triggers & Channel Coverage | Partial |
|
Runs start when a person or the customer's own code asks, through a task described in the browser, the CLI, the VS Code extension or an Agents API call over HTTP. Blackbox names no schedule, webhook or repository event that wakes an agent on its own. Sourceblackbox.ai/agents and docs.blackbox.airead 2026-09-29 |
|
| Model Flexibility & Routing | Full |
|
The Blackbox Router exposes more than 300 open and closed models through one OpenAI compatible endpoint, with the model named per request and provider routing. A multi agent task sets the model per agent, and Enterprise Inference runs the open weight model the customer chooses on reserved capacity. Sourceblackbox.ai and docs.blackbox.ai/api-reference/multi-agent-taskread 2026-09-29 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Blackbox documents an OpenAI compatible Inference API and a Dedicated API for its platform, and the pricing page includes an Agents API on the enterprise plan. The Agents API reference and the MCP server documentation that once sat at agent.blackbox.ai are no longer published. Sourcedocs.blackbox.ai and blackbox.ai/pricingread 2026-10-25 |
|
| Testing, Debugging & Optimization | Partial |
|
A multi agent task lets a team compare approaches, code quality and results across two to five agents on the same task, and Chairman LLM evaluates the implementations. That comparison covers one task at a time, and Blackbox names no harness to test, evaluate or regression check agent behavior over time. Sourcedocs.blackbox.ai/api-reference/multi-agent-task and blackbox.ai/agentsread 2026-09-29 |
|
| Browser & Computer Use | Not documented |
|
Agents work through repositories, sandboxes and APIs. Blackbox names no way for an agent to drive a browser or operate software through a graphical interface. Sourceblackbox.ai/agents and docs.blackbox.airead 2026-09-29 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
A per step model router in the Blackbox AI gateway sends each step of a coding agent session to a closed frontier model, a big open weight model or a small open weight model, depending on how hard that step is. Microsoft Decision-1 judges each step in about half a second, and Blackbox engineers configure the routing for each agent workload. In live Terminal-Bench 4.0 runs the router completed 65 percent of tasks at 38 percent lower cost per task than sending every step to Claude Opus.
Bears on: MCP / tool calling / API
View sourceBlackbox AI introduced a two-model orchestration architecture that combines GPT-5.6 Sol and Claude Opus 4.8. The system uses a sandbox-executing critic agent to evaluate responses and gate the final output, achieving improved accuracy on the Terminal-Bench v2.1 benchmark.
Bears on: Agent capability
View sourcePricing
Enterprise only: custom pricing on committed annual spend; no self serve plan
Committed annual spend drawn down by token usage across models.
Included quota
The committed balance covers token usage across 300+ models, with a dedicated forward deployed engineer and implementation at no extra cost.
What is public
Only an Enterprise plan is listed, priced on committed annual spend with discounts starting at ten percent on open weight models and five percent on closed models, improving with committed spend. No self serve tiers, free plan or trial are published.
Billing mechanics
The customer commits an annual spend, and the balance decreases as teams consume tokens across models, with alerts at 75 and 90 percent of the commitment.
Cost watchouts
Spend follows token usage against the commitment, so model choice and agent volume drive how fast the balance runs down.
Variable cost rationale
A committed balance is drawn down by usage, so cost is bounded by the commitment but depends on model mix and volume.
Additional watchouts
There is no self serve entry point: a sales conversation and a committed spend come first.
Overage / add-ons
Usage draws down the committed balance, with alerts at 75 and 90 percent; terms beyond the commitment are not published.
Sales call required
Yes, required for paid access
Free / trial
No free plan or trial; engagement starts with sales.
Lowest paid plan
None published; Enterprise only.
Commercial notes
Enterprise includes single tenant deployment, end to end encryption, contractual zero data retention, PII removal before closed models, SAML SSO, SCIM, RBAC, audit logs and the Agents API.
Key ambiguities
No dollar figures are published; cost depends on the committed spend and model mix.
Cancellation / refund
Annual committed spend agreed with sales; terms are not published.
Support SLA / resale
Priority compute/support on higher tiers; Enterprise adds SAML SSO, training opt-out by default, advanced security controls, and on-prem deployment
Missing data
Commitment minimums and per model token rates under the plan are not published.
Related vendors
- Cognition — Maker of Devin, an autonomous AI software engineer
- 10Web — Agentic website platform whose AI agents build, host and manage…
- AgentUI — Managed AI app builder for operations teams that generates and hosts…
- Aider — Open source, model agnostic terminal coding agent that edits your…
- Amp — Amp is a coding agent from Amp Frontier Corporation that runs each…
- Anthropic Claude Code — Anthropic's agentic coding system across terminal, desktop, IDE, web…
Alternatives to Blackbox AI
The closest documented capability profiles to Blackbox AI among coding agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- 10Web9.0 / 14Adds documented Prebuilt Agents, Templates & Packs
- CodeGPT10.0 / 14Adds documented Memory & State Persistence and Prebuilt Agents, Templates & Packs
- Zed10.0 / 14Adds documented Memory & State Persistence and Prebuilt Agents, Templates & Packs
- OpenCode10.5 / 14Adds documented Memory & State Persistence and Prebuilt Agents, Templates & Packs
- ThinkReview8.5 / 14Adds documented Memory & State Persistence and Prebuilt Agents, Templates & Packs
- Bubble10.0 / 14Adds documented Memory & State Persistence and Prebuilt Agents, Templates & Packs
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded